Abstract

The influence of the tropical Pacific sea surface temperature (SST) on the wintertime surface air temperature (SAT) in China is investigated using both the observational data and the output of coupled ocean–atmosphere numerical models during the period from 1960 to 2006. A singular value decomposition analysis (SVD) is applied between 500-hPa geopotential height (Z500) in the Northern Hemisphere and SST in the tropical Pacific Ocean to get the tropical Pacific SST-forced atmospheric patterns. The association of the SAT over China and the tropical Pacific SST is measured by calculating the temporal correlation coefficient (TCC) between the SAT and the expansion coefficient of the atmospheric component of the leading two SVD modes. Results show that the SAT over China is significantly correlated to the second SVD mode (SVD2). The SST component of SVD2 is characterized by negative tropical Pacific SST anomalies centered over the midequatorial Pacific Ocean. The atmospheric component of SVD2 (ASVD2) shares many similarities in spatial structures to the Arctic Oscillation (AO). The time variation of ASVD2, however, is found more closely correlated to the variation of SAT over China than the AO. When SVD2 is in its positive phase, the SAT over China tends to be warmer than normal. Further analysis indicates that the TCC between the SAT in China and ASVD2 is largely decreased after the long-term climate trend is removed. The time variability of the tropical Pacific SST-forced large-scale atmospheric patterns and its relationship to SAT are reasonably captured by the multimodel ensemble (MME) seasonal forecasts. An examination of the MME forecast skill indicates that ASVD2 contributes significantly to the TCC skill of MME forecasts.

1. Introduction

The Pacific–North American (PNA) pattern and the North Atlantic Oscillation (NAO) can significantly influence the atmospheric condition over the Northern Hemisphere (NH) extratropics (Trenberth et al. 1998; Hoerling et al. 2001). The PNA pattern, which is partly linked to the tropical Pacific SST associated with the El Niño–Southern Oscillation (ENSO), is believed to be the dominant source of seasonal forecast skill in wintertime (Shukla et al. 2000; Derome et al. 2001) that influences the weather and climate over the North Pacific and North American region. NAO, a dominant mode of variability in the North Atlantic region, is characterized by a meridional oscillation in atmospheric mass between centers of action near Iceland and over the subtropical Atlantic. The variability of NAO is associated with changes in the location of the storm track over the North Atlantic region as well as with precipitation and temperature anomalies over much of the North Atlantic and regions nearby (Rogers 1990; Hurrell 1995, 1996).

Although there exists some debate, many regard NAO as a regional display of the Arctic Oscillation (AO), which was defined first by Thompson and Wallace (1998). It is well accepted that the AO/NAO is, to a significant degree, an internal mode of variability of the atmospheric circulation. The spatial structure and amplitude of the AO/NAO can be well simulated in atmospheric general circulation models (AGCM) forced with fixed external forcing (Barnett 1985; Limpasuvan and Hartmann 1999). Recently, some observational and model studies indicated that part of NAO can be linked to the tropical forcing (Pozo-Vazquez et al. 2001; Wu and Hsieh 2004; Lin et al. 2005b; Li et al. 2006; Jia et al. 2009). For example, the results of Walter and Graf (2002) showed that when the NAO index is characterized by a pronounced decadal variability and by mainly positive values; the North Atlantic SST is strongly correlated to the regional atmospheric circulation in the North Atlantic sector. On the contrary, remote influence, in particular from the tropical Pacific region, becomes important when the NAO index is characterized by weak decadal variability. Pozo-Vazquez et al. (2001) and Wu and Hsieh (2004) demonstrated that a cold phase of ENSO has a robust remote impact over the North Atlantic and the corresponding sea level pressure (SLP) anomaly resembles the positive phase of NAO.

In the East Asian area, the East Asian winter monsoon (EAWM), which is mainly driven by the land–ocean thermal contrast, is one of the most active components of the global climate system in boreal winter (Ding 1994). The EAWM experiences a large year-to-year variability resulting from influence of multiple factors that make it very difficult to predict. Many efforts have been devoted to understanding the mechanisms and prediction of the variation of the winter monsoon circulation in East Asia (Chen et al. 2000; Wang et al. 2000; Zhang et al. 1996; Li 1990). Some studies showed that the EAWM usually becomes weaker than normal when there is a positive SST anomaly over the eastern tropical Pacific and vice versa. Moreover, the SST anomaly in the South China Sea that related to the EAWM can persist to the following summer and influences the precipitation in central-eastern China (Chen et al. 2000). Other atmospheric teleconnections may also play notable roles in the climate variability over the East Asian regions (Wang et al. 2007; Gong et al. 2001). Some studies showed that the EAWM tends to be weak during the positive phase of AO. The AO can influence the East Asian climate through the impact on the Siberian high (e.g., Gong et al. 2001) by modulating the East Asia jet stream (Yang et al. 2002; Branstator 2002; Gong and Ho 2003; Watanabe 2004), by teleconnection patterns (Li et al. 2006), or by modulating the quasi-biennial oscillation (QBO) wind in the stratosphere (Chen et al. 2003). The North Pacific Oscillation (NPO), a north–south seesaw of sea level pressure, is one of the most important teleconnection modes over the North Pacific region. Recent studies found that the EAWM tends to be weaker-than-normal when the NPO is strong (Guo and Sun 2004). Although many efforts have been devoted to understanding the mechanisms of the variation of the climate in East Asia, the question of which one of the proposed mechanisms is more essential to understand its variability still remains.

In previous studies, the ensemble forecasts produced under the second phase of the Canadian Historical Forecasting Project (HFP2) multimodel two-tier seasonal forecasting system conducted by the Canadian Meteorological Centre (Kharin et al. 2009) were used to examine the relationship between the tropical Pacific SST and large-scale atmospheric patterns over the extratropics (Lin et al. 2005a, 2008; Jia and Lin 2011), where a persistent SST anomaly was specified. They demonstrated that the surface air temperature (SAT) over China is significantly influenced by the SST-forced large-scale atmospheric patterns. This relationship could be potentially useful to improve the forecast skill of dynamical numerical models. The data used in the previous study are the output from four global atmospheric models for 33 yr from 1969 to 2001 with prescribed SST anomalies. There was, however, lack of air–sea interaction in the HFP2 system. Many studies demonstrated that air–sea coupling is important in climate variability. In this study, we examine the relationship between the SAT over China and the tropical Pacific SST-forced large-scale atmospheric patterns using the output of five air–sea coupled models for the period from 1960 to 2005 and focus on wintertime. We show that the SAT over China is significantly correlated to the second value decomposition analysis (SVD) mode (SVD2), which is to a large extent, the display of the long-term climate trends in the atmosphere and in the SST fields. The influence of the climate trend on the relationship between the SAT over China and the tropical Pacific SST-forced large-scale atmospheric patterns is examined. Furthermore, the contribution of the tropical Pacific SST-forced large-scale atmospheric patterns to the seasonal forecast skill of numerical models is evaluated. It is also of interest to compare the tropical Pacific SST-forced large-scale atmospheric patterns as well as their association to the climate over China between the output from the coupled ocean–atmosphere models and that from the HFP2. This comparison can help shed some light on the importance of the air–sea interaction in the coupled ocean–atmosphere numerical models.

The paper is organized as follows. In section 2, the data and model used in this study are described. The associations between the SAT over China and the large-scale atmospheric patterns obtained using a rotated empirical orthogonal function (REOF) are presented in section 3. Section 4 reveals the relationship between the SAT over China and the tropical Pacific SST-forced large-scale atmospheric patterns in the observations and in the multimodel ensemble (MME) seasonal forecasts. The importance of the long-term linear trend and SVD2 to the forecast skill of MME seasonal forecast skill is examined in section 5. In section 6, the tropical Pacific SST-forced large-scale atmospheric patterns and their relationship to the climate over China in the model forecasts from the ocean–atmosphere coupled models and from the HFP2 are compared, followed by the conclusion in section 6.

2. Data and models

The reanalyses of 500-hPa geopotential height (Z500), 200-hPa zonal wind, and the wind at 850 hPa from the National Centers for Environmental Predictions (NCEP)–National Center for Atmospheric Research (NCAR) reanalysis I (Kalnay et al. 1996), covering 47 yr from 1960 to 2006, are used in this study. The precipitation is from the National Oceanic and Atmospheric Administration (NOAA)’s Precipitation Reconstruction Dataset (PREC) (Chen et al. 2002). Two independent SAT datasets are used in this study. One is the Climate Research Unit (CRU) TS 2.1 dataset, a set of monthly averaged observed SAT over the land surface from the CRU) at the University of East Anglia, United Kingdom (Mitchell and Jones 2005) (http://www.cru.uea.ac.uk/cru/data/hrg.htm). The other observed dataset is the surface temperature of 160 meteorological stations in China for the same period. The data were collected and edited by China Meteorological Administration and were relatively homogeneously distributed, especially in eastern China (Wu et al. 2003). A comparison between these two observed datasets shows that they have quite similar structures according to the climatological mean and the standard deviation. The observed SST used in this study is the Met Office Hadley Centre Sea Ice and Sea Surface Temperature dataset (HadISST 1.1). It is a unique combination of monthly globally complete fields of SST and sea ice on a one-degree latitude–longitude grid (Rayner et al. 2003).

Also used in the present study are the outputs of seasonal ensemble forecast from five coupled models. Table 1 lists the acronyms of the institutions and models mentioned in the text. The coupled models are from CAWCR in APCC/CliPAS (Wang et al. 2009; Lee et al. 2010) and from CMCC-INGV, ECMWF, IFM-GEOMAR, MF, and UKMO in the ENSEMBLES project (hereafter, APCC-ENSEMBLES) (Weisheimer et al. 2009). None of the APCC-ENSEMBLES models have flux adjustments. A brief summary of the coupled models and their retrospective forecasts is presented in Table 2. In this study, we use the 1-month-lead seasonal forecasts starting from the first day of November. Therefore, the forecast seasons are the 3-month averages of December–February (DJF). The first month of integration is not used, as we are mainly interested in the forecast signal coming from air–sea coupling rather than that from the atmospheric initial conditions. The MME seasonal forecasts were made by the simplest approach of applying equal weights to the 3-month forecasts of all contributing models. All the calculations are performed using the forecast anomalies, computed for each individual model by removing its corresponding climatology. In this study, the winter of 1960 refers to the boreal 1960/61 winter.

Table 1.

Acronym names of institutions and models used in the text.

Acronym names of institutions and models used in the text.
Acronym names of institutions and models used in the text.
Table 2.

Description of the coupled models and their retrospective forecast used in this study.

Description of the coupled models and their retrospective forecast used in this study.
Description of the coupled models and their retrospective forecast used in this study.

3. Influence of atmospheric leading REOFs on wintertime SAT in China

We start by examining the influence of dominant atmospheric modes of variability in NH in wintertime. A REOFs analysis (Richman 1986) was conducted using the DJF seasonal-averaged Z500 of the NCEP–NCAR reanalysis north of 20°N for the period from 1960 to 2006. In the REOF analysis, 23 vectors are used for rotation. The first REOF (REOF1) and second REOF (REOF2) explain 10% and 8% of the data, respectively. The first two modes are display in Figs. 1a and 1b, respectively. Figure 1a shows a pattern of positive Z500 anomalies over the subtropical Pacific and over western Canada and negative Z500 anomalies over the North Pacific Ocean and southeastern United States. In Fig. 1b, the atmospheric pattern displays a negative anomaly centered over Iceland and a positive Z500 anomaly along the midlatitude North Atlantic. These two atmospheric patterns are consistent with the study of Lin et al. (2008, their Fig. 2) and are similar to the typical PNA and NAO structures defined by Wallace and Gutzler (1981) and Hurrell (1995), respectively. The principle component (PC) associated with the REOF1 (PC-REOF1) and REOF2 (PC-REOF2) are obtained by projecting the DJF anomaly of NCEP–NCAR Z500 to the REOF1 and REOF2 spatial patterns as shown in Figs. 1a and 1b, respectively. The temporal correlation coefficient (TCC) between the PC-REOF1 and the PNA index derived from the formula of Wallace and Gutzler (1981) (http://jisao.washington.edu/datasets/pna/) is 0.93 during the period under examination. The TCC between PC-REOF2 and the seasonally averaged station-based NAO index, which is the difference of normalized SLP between Ponta Delgada, Azores, and Stykkisholmur/Reykjavik, Iceland, is 0.92. Both TCCs obtained above pass the significance test with a significance level of 0.01 according to a Student’s t test.

Fig. 1.

First two modes of an REOF analysis of DJF 500-hPa height from NCEP–NCAR reanalysis from 1960 to 2006, represented as regression of DJF 500-hPa height onto the corresponding time expansion coefficients. The magnitudes correspond to one standard deviation of the time coefficient. Zero line is omitted. Contours with negative values are dashed. The contour interval is 10 m.

Fig. 1.

First two modes of an REOF analysis of DJF 500-hPa height from NCEP–NCAR reanalysis from 1960 to 2006, represented as regression of DJF 500-hPa height onto the corresponding time expansion coefficients. The magnitudes correspond to one standard deviation of the time coefficient. Zero line is omitted. Contours with negative values are dashed. The contour interval is 10 m.

The relationships between REOF1 and REOF2 and the SAT over China are evaluated by calculating the TCC between the PC-REOFs and the 160-station SAT for REOF1 (Fig. 2a) and REOF2 (Fig. 2b), respectively. Areas with a correlation significant at the 0.05 level or better, according to a Student’s t test, are shaded. The association between REOF1 and SAT over China is not very pronounced as can be seen from Fig. 2a, indicating that the impact of PNA in China is small. Only a limited area of significant correlation is observed over central China in the southern Gansu province. The TCC map for REOF2 is seen as more significant than REOF1 (Fig. 2b). Significant TCC appears in a large area over northeastern China as well as some areas over Xinjiang province. The relationships between the first two REOFs and the SAT over China are also examined using the CRU SAT and displayed in Figs. 2c and 2d for REOF1 and REOF2, respectively. The general features of the correlation maps are quite consistent with those obtained using the 160-station SAT dataset, although the TCC in Fig. 2b is slightly larger than that of Fig. 2d. In summary, the SAT in China over some areas of northeastern and northwestern China in wintertime is significantly affected by REOF2, whereas little relationship is observed between the SAT in China and REOF1, and this relationship can be observed using both the station data and the CRU data. In the following, only results obtained using the station data will be presented.

Fig. 2.

Temporal correlation coefficient between the 160-station SAT and the time series of (a) REOF1 and (b) REOF2. (c) As in (a), but for the CRU SAT; (d) as in (b), but for the CRU SAT. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

Fig. 2.

Temporal correlation coefficient between the 160-station SAT and the time series of (a) REOF1 and (b) REOF2. (c) As in (a), but for the CRU SAT; (d) as in (b), but for the CRU SAT. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

4. Tropical Pacific SST-forced large-scale atmospheric patterns and wintertime SAT in China

a. Observations

In a previous study, Jia and Lin (2011) examined the seasonality of the influence of the tropical Pacific SST-forced large-scale atmospheric patterns on the SAT over China using 33-yr observations and the output from the Canadian two-tier seasonal forecasting system of the HFP2. Their results indicate that the SAT over China can be significantly influenced by these large-scale atmospheric patterns. In this section, the relationship between the tropical Pacific SST-forced large-scale atmospheric patterns and the SAT over China is further examined using data of APCC-ENSEMBLES coupled ocean–atmosphere numerical models from 1960 to 2006 and focus will be given to wintertime. To get the large-scale atmospheric patterns related to the tropical Pacific SST forcing, an SVD analysis (Bretherton et al. 1992) is performed between the DJF seasonal-averaged NCEP–NCAR Z500 over the NH north of 20°N and the SST of the same season over the tropical Pacific (20°N–20°S, 120°E–90°W). The atmospheric and oceanic components of the first two SVDs are illustrated in Fig. 3. The atmospheric patterns of the leading two SVDs, depicted in Figs. 3a and 3b, represent the dominant-forced atmospheric patterns associated with the SST forcing in the tropical Pacific, which would represent the dominant source of seasonal forecast skill in the NH, especially in wintertime. The magnitudes of Z500 and SST correspond to one standard deviation of the expansion coefficients of the oceanic and atmospheric components of SVD, respectively. The first pair of SVD (SVD1) explains 76% of the total covariance between the Z500 and SST fields based on a squared covariance fraction, whereas that for the second pair of SVD2 is 12%. The correlations between the expansion coefficients of the Z500 (APC1 for SVD1 and APC2 for SVD2, hereafter) and the SST fields are 0.82 and 0.65 for SVD1 and SVD2, respectively, all significantly correlated to each other with a significance level of 0.05 according to a Student’s t test.

Fig. 3.

(a),(b) Observed DJF 500-hPa height and (c),(d) SST distributions of (left) SVD1 and (right) SVD2. The contour interval is (top) 5 m and (bottom) 0.15°C. Contours with negative values are dashed. Zero line is omitted in (a)–(c). The magnitude corresponds to one standard deviation of each time coefficient.

Fig. 3.

(a),(b) Observed DJF 500-hPa height and (c),(d) SST distributions of (left) SVD1 and (right) SVD2. The contour interval is (top) 5 m and (bottom) 0.15°C. Contours with negative values are dashed. Zero line is omitted in (a)–(c). The magnitude corresponds to one standard deviation of each time coefficient.

The atmospheric component of SVD1 (ASVD1) (Fig. 3a) is characterized by weak positive Z500 anomalies in the subtropical Pacific and over the northeastern North American region together with pronounced negative Z500 anomalies in the North Pacific Ocean. This atmospheric pattern shares many similarities in spatial structures to REOF1 as shown in Fig. 1a, as well as the PNA. The TCC between the APC1 (thick solid line in Fig. 4a) and the PC-REOF1 (dashed line in Fig. 4a) is 0.67, with a significance level of 0.05 according to a Student’s t test. On the other hand, ASVD2 (Fig. 3b) has negative Z500 anomalies around Greenland and the Bering Strait and positive anomalies along the extratropical North Atlantic and North Pacific, which is similar to REOF2 shown in Fig. 1b. The corresponding TCC between the time series of ASVD2 (APC2) (thick solid line in Fig. 4b) and PC-REOF2 (dashed line in Fig. 4b) reaches 0.81, again, with a significance level of 0.05 according to a Student’s t test. The associated SST component of SVD1 (SSVD1) (Fig. 3c) represents a typical El Niño structure with positive SST anomalies dominant over the eastern tropical Pacific. The expansion coefficient of SSVD1 (SPC1) is highly correlated to the Niño-3.4 index. The TCC between SPC1 and the Niño-3.4 index obtained from the NOAA Climate Prediction Center (CPC) that is the average of the SST anomaly over the eastern-central tropical Pacific (5°N–5°S, 170°–120°W) is 0.97 during the period under examination. The SST distribution for SVD2 features negative SST anomalies over the tropical Pacific with a maximum negative SST anomaly over the central equatorial Pacific Ocean, consistent with Jia and Lin (2011).

Fig. 4.

(a) The time series associated with REOF1 (thick solid line), APC1 (dashed line) for the observations and APC1 (thin line with clear box) for the MME seasonal forecasts. (b) As in (a), but for REOF2 and APC2.

Fig. 4.

(a) The time series associated with REOF1 (thick solid line), APC1 (dashed line) for the observations and APC1 (thin line with clear box) for the MME seasonal forecasts. (b) As in (a), but for REOF2 and APC2.

To assess the association between the SAT over China and the leading SST-forced atmospheric patterns, the TCCs between the 160-station SAT and APC1 and APC2 are calculated and depicted in Figs. 5a and 5b, respectively. Figure 5a shows significant TCC appearing over some areas of central China mainly located over the Yellow River basin. Compared to SVD1, the SAT over China is much more significantly correlated to SVD2. It can be seen from Fig. 5b that significant positive correlation occurs over most of northern China. The lower reaches of Yangtze River and parts of southwestern China, such as southeastern Tibet, are also covered by significant positive TCC. The above result indicates that when SVD2 is in its positive phase, which is represented by a cooling SST anomaly over the central tropical Pacific in the ocean and an AO-like atmospheric response in the atmosphere, the SAT over China in wintertime tends to be warmer than normal. A comparison between Figs. 5b and 2d shows that, although APC2 and PC-REOF2 are highly correlated, the associated TCC for APC2 is much more significant than that for PC-REOF2, indicating that the SAT in China is more affected by the atmospheric pattern obtained using the above SVD analysis. The regression of Z500 to APC2 shows pronounced negative anomalies over the Aleutian Islands and significant positive anomalies over the subtropical North Pacific (not shown). The easterly winds along the south flank of the positive Z500 anomalies flow all the way westward to the East Asian continent and bring warm air from the ocean to the continent resulting in a warmer-than-normal winter over there.

Fig. 5.

Temporal correlation coefficient between the observed SAT and the (a) APC1 and (b) APC2 for the period from 1960 to 2006. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

Fig. 5.

Temporal correlation coefficient between the observed SAT and the (a) APC1 and (b) APC2 for the period from 1960 to 2006. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

To have a better idea of to what extent the tropical Pacific SST-forced leading atmospheric patterns impact the climate over China, we examine the climatological mean and standard deviation of SAT in DJF for observations (Fig. 6). In the observations, the climatological DJF SAT increases from north to south while the values of the standard deviation decrease from north to south with maximum values over the northwest and northeast regions. The regressions of SAT to APC1 and APC2 are presented in Figs. 7a and 7b, respectively. It is seen that the distributions of the regression patterns are similar to the correlation map as shown in Fig. 5. The largest value appears over northeast and northwest China with values higher than 0.7°C, as shown in Fig. 7b. The ratios of the regression patterns to the standard deviation of the DJF mean SAT are presented in Figs. 7c and 7d. It is seen that the ratio of APC1 over central-east China reaches 0.3. The ratio for APC2 is higher than that of APC1 with most of northern China higher than 0.3.

Fig. 6.

(a) Time average and (b) standard deviation of the DJF SAT over China. The contour interval is 5 in (a) and 0.2 in (b). Areas greater than zero and 0.8°C are shaded in (a) and (b).

Fig. 6.

(a) Time average and (b) standard deviation of the DJF SAT over China. The contour interval is 5 in (a) and 0.2 in (b). Areas greater than zero and 0.8°C are shaded in (a) and (b).

Fig. 7.

The regression of SAT to (a) APC1 and (b) APC2. The ratio of the regression of SAT to the standard deviation of the SAT for (c) SVD1 and (d) SVD2. The contour interval is 0.2°C in (a) and (b) and 0.2 for (c) and (d). Areas with a significance level of 0.05 according to a Student’s t test are shaded in (a) and (b), and areas greater than 0.2 are shaded in (c) and (d).

Fig. 7.

The regression of SAT to (a) APC1 and (b) APC2. The ratio of the regression of SAT to the standard deviation of the SAT for (c) SVD1 and (d) SVD2. The contour interval is 0.2°C in (a) and (b) and 0.2 for (c) and (d). Areas with a significance level of 0.05 according to a Student’s t test are shaded in (a) and (b), and areas greater than 0.2 are shaded in (c) and (d).

Figure 4 shows that both APC2 and PC-REOF2 display a distinctive upward trend during the period under examination. The linear trends of APC2 and PC-REOF2 are 0.50 and 0.37 decade−1 during this period, respectively. To better understand the SVD2 mode, the regression of SST onto the SPC2 and the climate trend in the SST field during the same period over the tropical Pacific Ocean are illustrated in Figs. 8a and 8b, respectively. The linear trend of SST shows cooling anomalies along the equatorial Pacific and central-north of the tropical Pacific, consistent with previous studies (Solomon and Newman 2012). The linear trend SST pattern is also to a good extent similar to Fig. 8a. We also examine the long-term linear trend in Z500 during the period from 1960 to 2006 in the observations. Figure 9a shows the long-term climate trend of Z500 of the NCEP–NCAR reanalysis over the NH. As is seen, the linear trend pattern in Z500 has negative anomalies with one center near Greenland and the other near the Bering Strait. Positive Z500 anomalies can be seen along the midlatitude North Atlantic. This linear trend pattern shares some similarities to the spatial distribution of ASVD2 as shown in Fig. 3b. The above results indicate that SVD2, to a large extent, is the display of the long-term climate trends in the atmosphere and in the SST fields. The long-term precipitation trends (Fig. 9b) are dominated by pronounced dry anomalies over the central-western tropical Pacific, consistent with the cooling SST trends shown in Fig. 8b. A relatively weaker positive precipitation anomaly center can also be found over the central-eastern tropical Pacific. Using a simple atmospheric model, Jia et al. (2009) examined the atmospheric response to the thermal forcing over different locations of equatorial Pacific and found that a negative thermal forcing over the western-central tropical Pacific is effective in generating an AO-like (or NAO-like) atmospheric response.

Fig. 8.

(a)The regression of SST onto the normalized time series associated with SPC2 and (b) the linear trend of the global SST. The contour interval is 0.1°C for (a) and 0.1°C decade−1 for (b). Contours with negative values are shaded.

Fig. 8.

(a)The regression of SST onto the normalized time series associated with SPC2 and (b) the linear trend of the global SST. The contour interval is 0.1°C for (a) and 0.1°C decade−1 for (b). Contours with negative values are shaded.

Fig. 9.

The long-term linear climate trend during the period from 1960 to 2006 in (a) Z500 and (b) precipitation in the observations. The contour interval is 5 m decade−1 for (a) and 0.3 mm decade−1 for (b).

Fig. 9.

The long-term linear climate trend during the period from 1960 to 2006 in (a) Z500 and (b) precipitation in the observations. The contour interval is 5 m decade−1 for (a) and 0.3 mm decade−1 for (b).

To demonstrate the impact of climate trend on the relationship between the tropical Pacific SST anomaly and the SAT over China, the long-term linear trends are removed from the PC-REOFs and APCs before TCC is calculated. The TCC between the detrended APC1 and PC-REOF1 is 0.72 and that between the detrended APC2 and PC-REOF2 is 0.70, respectively, with a significance level of 0.05 according to a Student’s t test. The TCC between the detrended PC-REOFs and the SAT over China are calculated and illustrated in Figs. 10a and 10b, respectively, while that for APC1 and APC2 are illustrated in Figs. 10c and 10d, respectively. The correlations for the detrended PC-REOF1 and APC1 (Figs. 10a,c) are quite weak with significant correlations only over a very limited area in central China. The TCC values for the detrended PC-REOF2 and APC2 (Figs. 10b,d) are found substantially decreased compared to their counterparts in the original data (Figs. 2b, 5b). There is almost no significant TCC between the SAT over China and the detrended PC-REOF2 and detrended APC2. The above analysis indicates that the long-term climate trend significantly influences the relationship between the tropical Pacific SST-forced large-scale atmospheric patterns and the SAT over China, especially for the AO-like atmospheric pattern. In a previous study, Gong et al. (2001) investigated the relationship between the AO and variability of the East Asian winter monsoon. They pointed out that there are significant out-of-phase relationships between the AO and the East Asian winter monsoon. The AO and the temperature of eastern China are correlated with a TCC of 0.34 during the period from 1951 to 1999. However, when the linear trend is removed from the data, the coefficient between the AO index and the surface temperature over eastern China is only 0.17, which is no longer significant (Gong et al. 2001) and is consistent with our study.

Fig. 10.

Temporal correlation coefficient between the observed SAT and the detrended time series associated with (a) REOF1, (b) REOF2, (c) APC1, and (d) APC2. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

Fig. 10.

Temporal correlation coefficient between the observed SAT and the detrended time series associated with (a) REOF1, (b) REOF2, (c) APC1, and (d) APC2. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

b. MME seasonal forecasts

For a seasonal forecast to be skillful, it is important for numerical models to capture the relationship between the atmospheric response and SST forcing as shown in the last section. In this section, the large-scale atmospheric patterns and associated tropical Pacific SST forcing in the MME seasonal forecast are examined using the output from APCC-ENSEMBLES. A similar SVD analysis as we did for the observations is applied but using the ensemble 1-month-lead forecast of seasonal-mean Z500 and SST during the same period from 1960 to 2006. Figure 11 shows the spatial structures of the first two pairs of SVD modes in the MME seasonal forecast. The atmospheric component of SVD1 and SVD2 are shown in Figs. 11a and 11b, respectively, while the corresponding oceanic components are shown in Figs. 11c and 11d. The magnitude of the atmospheric component of SVD1 in the MME Z500 forecasts (Fig. 11a) is much stronger than its counterpart in the observations (Fig. 3a). The spatial structure of the atmospheric component of SVD1 in the MME Z500 forecasts bears some similarities to ASVD1 over the North Pacific and North American region. However, obvious disagreement can also be found over the North Atlantic and European sector where a pronounced negative Z500 anomaly is seen over the extratropical region. Compared to ASVD1, which is more localized over the North Pacific–North American region, the atmospheric component of SVD1 in the MME Z500 forecasts is more global in spatial structure. The spatial distribution of the atmospheric component of SVD2 in the MME seasonal forecasts (Fig. 11b) shares some similarities to the AO pattern and ASVD2 in the observations. However, the atmospheric component of SVD2 in the MME Z500 forecasts has more weight over the high-latitude North Pacific than ASVD2 in the observations (Fig. 3b), and the dipole structure in the North Atlantic sector is not as clear. Compared to the atmospheric component, there are more similarities of the spatial structures of the oceanic component of SVD1 and SVD2 in the MME seasonal forecast and the observations. The ENSO signal is very clear in the oceanic component of SVD1 in the MME seasonal forecasts, as can be seen from Fig. 11c. The negative SST anomalies along the central tropical Pacific in the SST component of SVD2 are also well represented in the oceanic component of SVD2 in the MME seasonal forecasts (Fig. 11d).

Fig. 11.

As in Fig. 3, but for the MME seasonal forecasts. The contour interval is (a) 25 and (b) 10 m and (c),(d) 0.15°C. Zero line is omitted in (a)–(c). The magnitude corresponds to one standard deviation of each time coefficient.

Fig. 11.

As in Fig. 3, but for the MME seasonal forecasts. The contour interval is (a) 25 and (b) 10 m and (c),(d) 0.15°C. Zero line is omitted in (a)–(c). The magnitude corresponds to one standard deviation of each time coefficient.

In a previous study (Jia and Lin 2011), the influence of the tropical Pacific SST-forced large-scale atmospheric patterns on the climate over China is examined using the MME seasonal forecasts from HFP2 during the period from 1969 to 2001. The SST forcing used for the seasonal forecasts in HFP2 was the sum of the SST anomaly of the month prior to the forecast period, persisted through the forecast period, and the monthly varying climatological SST. Results show that the SVD1 pattern of the atmospheric component in the observations is reasonably reproduced by the numerical models in DJF. The SVD2 pattern, however, is found to be far from the observations that appear to be a wave train pattern over the North Pacific–North American regions.

To examine the performance of the MME seasonal forecasts in capturing the time evolution of the tropical Pacific SST-forced large-scale atmospheric patterns, the time variability of the atmospheric component of SVD1 in the MME Z500 forecasts (AMPC1) and SVD2 (AMPC2), and the time variability of the oceanic component of SVD1 and SVD2 is examined and compared to the observations. The TCCs for individual numerical models, as well as the MME seasonal forecasts, are calculated and illustrated in Table 3. The performance of individual numerical models is also examined to see the model differences of the MME seasonal forecasts. It can be seen that the time evolution of both the atmospheric and the oceanic components of SVD1 and SVD2 in the observations and in the forecasts are significantly correlated to each other for most numerical models. The only exception for TCC that cannot pass the significance test is the TCC between the APC1 and AMPC1 for MF and the TCC between the time series associated with the oceanic component of SVD1 in the observations and in UKMO. The TCC between APC1 in the observations and AMPC1 in the MME seasonal forecasts is 0.73, while the TCC between APC2 and AMPC2 is 0.45. The correlations of the time series associated with the oceanic component between the observations and the MME SST forecasts are 0.63 and 0.60 for SVD1 and SVD2, respectively. The above results indicate that although the spatial structures of the atmospheric response to the tropical Pacific SST forcing in the MME forecasts are to some extent distorted compared to the observations, the time evolution of both the atmospheric and oceanic components of the SVDs are reasonably well captured by the MME forecasts. Also noticed from Table 3 is that the MME prediction skill is generally better than the average forecast skill (AVE) of all individual models, indicating that the MME method is a valuable approach for reducing errors and quantifying forecast uncertainty due to model formulation. The above results show that the structure of the atmospheric response to the tropical Pacific SST forcing in both the one-tier and two-tier numerical models has some bias compared to the observations, especially for high-order SVD modes. The time series associated with these patterns are found to be well correlated to those in the observations to a certain extent.

Table 3.

Correlations between the atmospheric and oceanic expansion coefficients of SVD1 and SVD2 in the observations and in the numerical models’ seasonal forecasts. The correlations with a significance level of 0.05 according to a Student’s t test are set in bold.

Correlations between the atmospheric and oceanic expansion coefficients of SVD1 and SVD2 in the observations and in the numerical models’ seasonal forecasts. The correlations with a significance level of 0.05 according to a Student’s t test are set in bold.
Correlations between the atmospheric and oceanic expansion coefficients of SVD1 and SVD2 in the observations and in the numerical models’ seasonal forecasts. The correlations with a significance level of 0.05 according to a Student’s t test are set in bold.

The TCC between the model forecast SAT and the AMPC1 and AMPC2 are calculated and depicted in Figs. 12a and 12b, respectively. It is seen that the variation of SAT over China is more correlated to AMPC2 than AMPC1, consistent with the observations. The TCC map between the AMPC2 and SAT is found to be similar to Fig. 5b over northeastern China. However, the TCC between the AMPC1 and the SAT is different from Fig. 5a. The significant TCC in Fig. 12a appears over northwest and southeast China in the MME forecasts. The AMPC1 and AMPC2 of the MME seasonal forecasts are depicted as thin line with the open box in Figs. 4a and 4b, respectively. The AMPC2 also displays a pronounced upward trend with the linear trend at 0.32 decade−1. To examine the influence of the long-term linear trend to the relationship between the SAT over China and the SST-forced atmospheric patterns in the MME seasonal forecasts, the TCCs between SAT and the detrended AMPC1 and AMPC2 are also examined and presented in Figs. 12c and 12d, respectively. It can be seen that the TCC for SVD1 is still weak, as before, and the TCC for SVD2 is largely decreased. Only a very small area of northeastern China has significant correlations. The results indicate that the connection between the SAT over China and AMPC2 are also significantly influenced by the long-term linear trend, consistent with the observations.

Fig. 12.

Temporal correlation coefficient between the MME SAT and the (a) APC1 and (b) APC2 in the MME seasonal forecasts. (c) As in (a), but for the detrended APC1; (d) as in (b), but for the detrended APC2. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.2. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

Fig. 12.

Temporal correlation coefficient between the MME SAT and the (a) APC1 and (b) APC2 in the MME seasonal forecasts. (c) As in (a), but for the detrended APC1; (d) as in (b), but for the detrended APC2. Absolute values with TCCs greater than 0.2 are plotted. The contour interval is 0.2. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

5. Forecast skill of the MME seasonal forecasts

In previous sections, we showed that the SAT over China is significantly influenced by ASVD2. The time variability of the tropical Pacific SST-forced large-scale atmospheric patterns and its relationship to SAT over China are reasonably well captured by the APCC-ENSEMBLES coupled ocean–atmosphere numerical models involved in this study. In this section, the influence of the climate trend and the tropical Pacific SST-forced large-scale atmospheric patterns on the forecast skill of MME is examined. Figure 13a shows the forecast skill of numerical models as measured by TCC. The TCC is calculated between the observations and the MME seasonal forecasts during the period from 1960 to 2006 at each grid point. Areas with a significance level of 0.05, according to a Student’s t test are shaded. It is observed from Fig. 13a that the MME forecasts have statistically significant forecast skill over large parts of northeastern China and the Xinjiang province and some areas of southwestern China, similar to the spatial distribution of the TCC map shown in Fig. 5b. To see the influence of the climate trend on the MME forecast skill, the TCC between the detrended SAT in the MME forecasts and in the observations are also calculated and presented in Fig. 13b. It is found that the TCC skill of the MME forecasts is largely decreased, indicating that most of the MME forecast skill shown in Fig. 13a comes from the contribution of the long-term linear trends in the data.

Fig. 13.

Temporal correlation coefficient between the observed SAT and the (a) MME SAT forecast, (b) MME SAT forecast after removing the linear trend, (c) MME SAT forecast after removing the variability associated with APC1, and (d) MME SAT forecast after removing the variability associated with APC2. Absolute values with TCCs greater than 0.3 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

Fig. 13.

Temporal correlation coefficient between the observed SAT and the (a) MME SAT forecast, (b) MME SAT forecast after removing the linear trend, (c) MME SAT forecast after removing the variability associated with APC1, and (d) MME SAT forecast after removing the variability associated with APC2. Absolute values with TCCs greater than 0.3 are plotted. The contour interval is 0.1. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

To see the influence of the tropical Pacific SST-forced large-scale atmospheric patterns on the MME forecast skill, the linear trends associated with the SVD1 and SVD2 are removed by means of a linear regression. The TCC between the MME SAT forecast after removing the variability associated with the SVD1 and SVD2 and the observations is calculated and presented in Figs. 13c and 13d, respectively. Figure 13c shows that the TCC skill of the MME forecasts without the signal related to SVD1 is similar to the original data as shown in Fig. 13a. However, decrease over some areas, for example, the Mongolia area, northwestern China, can also be noticed. In contrast to SVD1, the TCC skill of the MME forecasts without the signal of SVD2 is largely decreased as can be seen from Fig. 13d. The significant TCC skill area shown in Fig. 13a is lost over many areas as can be seen from Fig. 13d, indicating the importance of the SVD2 to the MME SAT forecast skill.

6. The climate trends in the one-tier and two-tier forecasting systems

There exist two commonly used frameworks to do seasonal forecasts using numerical models (i.e., one-tier and two-tier frameworks). The two-tier model-based dynamical seasonal forecasts consist of two steps. The global sea surface temperature anomalies (SSTAs, the first tier) are first predicted by either a coupled GCM or a statistical model. Then the preforecasted SSTAs are used as boundary conditions for an AGCM to make forecasts for the coming season in the second step (the second tier). In contrast, the one-tier dynamical seasonal forecasts use only one coupled ocean–atmosphere model (CGCM) that contains interacting physics between the atmosphere and ocean. Previous studies showed that the CGCMs had considerable systematic errors in simulating the status of the tropical ocean and atmosphere, although it is closer to the nature than the two-tier model forecasting system. As the systematic forecast errors could potentially cause bias in the global teleconnections that are associated with equatorial SSTA, the two-tier approach system is demonstrated to have an obvious advantage over the direct use of the CGCMs for the extratropics. However, with the development of coupled climate models in the past decades, it might be possible for the CGCMs to improve their ability in capturing characteristics of the tropical forcing and extratropical teleconnections. Recently, some studies have showed that it is important to take into account local monsoon–warm pool ocean interactions in the seasonal forecasts (Wang et al. 2003; Wu and Kirtman 2005; Kumar et al. 2005). For example, Kumar et al. (2005) shows the superior forecast ability of a one-tier system over a two-tier system in predicting Indian subcontinent rainfall. Wu and Kirtman (2005) reported that local coupled air–sea feedback over the Indian Ocean plays an important role in simulating a proper monsoon–ENSO relationship in the dynamical model. However, most previous studies concentrated on a comparison between the two forecasting systems over the tropical region and on an interannual time scale. The difference between one-tier and two-tier forecast systems over the extratropics and on a longer time scale, however, has not yet been well reported.

As we mentioned before, in a previous study, the output from HFP2, a two-tier forecasting system, was used to examine the influence of the tropical Pacific SST-forced large-scale atmospheric patterns to the climate over the extratropics. In the HFP2 forecasting system, the SST anomaly from the month prior to the expected forecast period is added to the climatological annual cycle of SST and persists through the forecast period. One purpose of this study is to compare the MME seasonal forecast from APCC-ENSEMBLES coupled ocean–atmosphere models to that from HFP2 in capturing the tropical Pacific SST-forced atmospheric patterns as well as their relationships to the climate over China to shed some light on the importance of the air–sea interaction in numerical models. As climate trends are sensitive to the time period under examination (e.g., Figs. 9b, 14a), in this section the common period between the two forecasting systems from 1969 to 2001 is examined.

Fig. 14.

The long-term linear climate trend during the period from 1969 to 2001 for (a) observations, (b) APCC-ENSEMBLES, and (c) HFP2 precipitation forecasts. Units are mm decade−1.

Fig. 14.

The long-term linear climate trend during the period from 1969 to 2001 for (a) observations, (b) APCC-ENSEMBLES, and (c) HFP2 precipitation forecasts. Units are mm decade−1.

In section 4, it is demonstrated that, compared to SVD1, the SAT over China is more significantly correlated to SVD2, which is, to a large extent, the display of the long-term climate trends in the atmosphere and in the ocean fields. Therefore, it is important for model forecasts to have a reasonable representation of the long-term linear trend in the associated fields to capture the characteristics of SVD2. We start by examining and comparing the behaviors of the APCC-ENSEMBLES and HFP2 models in forecasting the linear trends of the tropical Pacific forcing for 33 yr from 1969 to 2001. The climate trends of precipitation in the observations and in the model forecasts are presented in Fig. 14. As is seen, the climate trend of the precipitation in the observations (Fig. 14a) is dominated by U-shaped negative anomalies centered over the central-western tropical Pacific. Positive trend anomalies can be noticed over the central-eastern tropical Pacific. The climate trends of the forecast precipitation from APCC-ENSEMBLES (Fig. 14b) and HFP2 (Fig. 14c) have similar spatial distributions over the tropical Pacific region. The climate trends of the precipitation in the HFP2 forecasts have relatively larger amplitudes compared to those in the APCC-ENSEMBLES forecasts. The U-shape negative trend anomalies over the western tropical Pacific and positive trend anomalies over the eastern tropical Pacific appearing in the observations can be noticed in both Figs. 14b and 14c. However, a further examination shows that the positive precipitation trend anomalies shift southward in the HFP2 forecasts compared to its counterpart in the observations, leaving the precipitation trends dominated by negative band along the equatorial Pacific. The pronounced positive precipitation trends around 10°N in the HFP2 forecasts do not appear in the observations. Generally speaking, compared to HFP2 forecasts, the climate trends of the precipitation in APCC-ENSEMBLES forecasts are closer to those in the observations at least during the period under examination.

The climate trends of Z500 in the observations and in the model forecasts are also examined and presented in Fig. 15. The Z500 trends in the observations (Fig. 15a) display positive anomalies over the tropics. Three significant positive Z500 trends centered over the NH extratropics can be observed with one over Europe, one over East Asia, and one over the North American continent. Two weak but significant negative Z500 trend centers can also be noticed with one around the Bering Strait and the other around Newfoundland. The linear trends of Z500 in the APCC-ENSEMBLES model forecasts (Fig. 15b) are much weaker in magnitudes compared to the observations, while the spatial distribution of the trend pattern is quite similar to that in the observations. The two negative Z500 trend centers, however, are not well captured by the APCC-ENSEMBLES model forecasts Fig. 15b. It needs to be kept in mind that the MME forecasts only represent the signal from the external forcings while the observations are the results of both the boundary forcings and the internal dynamics of the atmosphere. The climate trends of Z500 in the HFP2 forecasts also have three positive centers in the NH extratropics. However, the observed positive trend centered over East Asia shifts eastward to the central North Pacific, as can be seen from Fig. 15c.

Fig. 15.

The long-term linear climate trend of Z500 during the period from 1969 to 2001 for (a) observations, (b) APCC-ENSEMBLES, and (c) HFP2. The contour interval is 5, 1.5, and 3 m for (a),(b), and (c), respectively. Units are m decade−1.

Fig. 15.

The long-term linear climate trend of Z500 during the period from 1969 to 2001 for (a) observations, (b) APCC-ENSEMBLES, and (c) HFP2. The contour interval is 5, 1.5, and 3 m for (a),(b), and (c), respectively. Units are m decade−1.

The above results indicate that the observed climate trends in both the precipitation and Z500 fields can be reasonably captured by both the APCC-ENSEMBLES and HFP2 model forecasts. However, the APCC-ENSEMBLES forecasts can better capture the climate trends than the HFP2 model forecasts in both the tropics and NH extratropics, at least for the variables and period under examination. To compare the numerical models’ ability in capturing the association between the tropical Pacific SST-forced large-scale atmospheric patterns and the climate over China, the TCC between the forecast SAT over China and the APC2 from both APCC-ENSEMBLES and HFP2 are examined and presented in Fig. 16. It can be seen that the TCC is much more significant in the APCC-ENSEMBLES model forecast (Fig. 16a) compared to that in the HFP2 model forecasts (Fig. 16b), indicating the superiority of using one-tier forecasting systems over two-tier forecasting systems for capturing the relationship between the SVD2 and climate over China.

Fig. 16.

Temporal correlation coefficient between the model forecast SAT and the APC2 for (a) APCC-ENSEMBLES and (b) HFP2 for the period from 1969 to 2001. The contour interval is 0.2. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

Fig. 16.

Temporal correlation coefficient between the model forecast SAT and the APC2 for (a) APCC-ENSEMBLES and (b) HFP2 for the period from 1969 to 2001. The contour interval is 0.2. Zero line is omitted. Contours with negative values are dashed. Areas with a significance level of 0.05 according to a Student’s t test are shaded.

7. Summary and discussion

In this study, the relationship between the tropical Pacific SST-forced large-scale atmospheric patterns and SAT over China in wintertime is investigated using both the observational data and the MME seasonal forecast from coupled ocean–atmosphere numerical models during the period from 1960 to 2006. The tropical Pacific SST-forced large-scale atmospheric patterns are obtained by applying an SVD analysis between the DJF Z500 over the NH and simultaneous tropical Pacific SST, which is the main source of seasonal predictive skill. The influence of these large-scale atmospheric patterns on the SAT over China is measured by calculating the TCC between the time series associated with the leading two pairs of SVD modes and the SAT over China. The atmospheric components of the leading two SVD modes are found to have PNA- and AO-like patterns, respectively. The corresponding SST pattern for SVD1 is a clear ENSO signal with a pronounced positive SST anomaly lying over the eastern tropical Pacific, and the associated SST distribution for SVD2 is dominated by negative SST anomalies centered over the midequatorial Pacific. The SAT over China is found to be significantly influenced by the atmospheric component of SVD2 where positive correlation occurs over most of China. The result indicates that when SVD2 is in its positive phase, which is represented by a cooling SST anomaly centered over the midequatorial Pacific in the ocean and an AO-like atmospheric response in the atmosphere, the SAT over a large area of China tends to be warmer than normal. The MME seasonal forecasts in general reasonably simulate the relationship between the tropical Pacific SST forcing and the atmospheric response as in the observations. Although the spatial structures of the leading SST-forced atmospheric patterns are somehow distorted in the MME seasonal forecasts, the time variability of the atmospheric patterns is well captured by both the individual numerical models and the MME seasonal forecasts. The contribution of the tropical Pacific SST-forced large-scale atmospheric patterns to the forecast skill of MME forecasts are examined by calculating the TCC between the SAT in the observations and in the MME seasonal forecasts without the signal related to SVDs. It shows that the TCC skill of the MME seasonal forecasts was largely decreased after removing the signal related to SVD2, suggesting the importance of the SVD2 to the MME forecast skill. The influence of the long-term climate trends to the close relationship between the tropical Pacific SST-forced large-scale atmospheric patterns and SAT over China is also examined in this study. The TCC between the SAT over China and APC2 is largely decreased after the long-term linear trend is removed from the data. The TCC skill of the MME forecast is almost lost after the long-term linear trend is removed from the data. Previous studies demonstrated the superior prediction ability of a one-tier system over the corresponding two-tier system in predicting the SST and precipitation over tropics on the interannual time scale as the former includes ocean–atmosphere interaction processes. In this study, the performances of one-tier and two-tier approaches are compared to each other in capturing the climate trends as well as the association between the tropical Pacific SST-forced large-scale atmospheric patterns and the climate over China. It is also found that there are some advantages in using one-tier forecasting systems than two-tier forecasting systems.

In this study, we demonstrated that the SAT over China is significantly correlated to SVD2, which is to a large extent the display of the long-term climate trends in the atmosphere and in the SST fields for both the observations and MME forecasts. In this study, we show that the association between the SAT in China and SVD2 is mainly due to the trends. This indicates that a significant part of the SAT trends in China is caused by the response to the trends of the SVD2 mode. The total trends of SAT in China should be the sum of trends of different origins. For example, the direct effect of greenhouse gases may cause a SAT trend in China, which could be either positive or negative. Finally, it should be kept in mind that the leading atmospheric patterns examined in this study are mainly forced by the tropical Pacific SST forcing. It is possible that other external forcings, such as the SST anomalies in other ocean basins, for example, the Indian Ocean, can also influence the East Asian climate and may not be well represented by these modes.

Acknowledgments

This research was jointly funded by National Natural Sciences Foundation of China (Grants 41105037 and 91337216) and by the Natural Science and Engineering Research Council of Canada (NSERC). Jia is also supported by the Fundamental Research Funds for the Central Universities of China (Grant 2012XZZX012). We acknowledge Dr. June-Yi Lee for providing the model data for analysis in this study. We are grateful to the four reviewers for their helpful suggestions on improving our paper.

REFERENCES

REFERENCES
Alessandri
,
A.
,
A.
Borrelli
,
S.
Masina
,
A.
Cherchi
,
S.
Gualdi
,
A.
Navarra
, and
P.
Di Pietro
,
2010
:
The INGV–CMCC seasonal prediction system: Improved ocean initial conditions
.
Mon. Wea. Rev.
,
138
,
3634
3649
.
Balmaseda
,
M.
,
A.
Vidard
, and
D. L. T.
Anderson
,
2008
:
The ECMWF Ocean Analysis System: ORA-S3
.
Mon. Wea. Rev.
,
136
,
3018
3034
.
Barnett
,
T.
,
1985
:
Variations in near-global sea level pressure
.
J. Atmos. Sci.
,
42
,
478
501
.
Branstator
,
G.
,
2002
:
Circumglobal teleconnections, the jet stream waveguide, and the North Atlantic Oscillation
.
J. Climate
,
15
,
1893
1910
.
Bretherton
,
C.
,
C.
Smith
, and
J. M.
Wallace
,
1992
:
An intercomparison of methods for finding coupled patterns in climate data
.
J. Climate
,
5
,
541
560
.
Chen
,
M.
,
P.
Xie
,
J.
Janowiak
, and
P.
Arkin
,
2002
:
Global land precipitation: A 50-yr monthly analysis based on gauge observations
.
J. Hydrometeor.
,
3
,
249
266
.
Chen
,
W.
,
H. F.
Graf
, and
R. H.
Huang
,
2000
:
The interannual variability of East Asian winter monsoon and its relation to the summer monsoon
.
Adv. Atmos. Sci.
,
17
,
46
60
.
Chen
,
W.
,
M.
Takahashi
, and
H.-F.
Graf
,
2003
:
Interannual variations of stationary planetary wave activity in the northern winter troposphere and stratosphere and their relations to NAM and SST
.
J. Geophys. Res.
,
108
, 4797, doi:10.1029/2003JD003834.
Collins
,
W.
, and
Coauthors
,
2008
: Evaluation of the HADGEM2 model. Met Office Tech. Note 74, 47 pp. [Available online at http://www.metoffice.gov.uk/media/pdf/8/7/HCTN_74.pdf.]
Daget
,
N.
,
A. T.
Weaver
, and
M. A.
Balmaseda
,
2009
:
Ensemble estimation of background-error variances in a three-dimensional variational data assimilation system for the global ocean
.
Quart. J. Roy. Meteor. Soc.
,
135
,
1071
1094
.
Derome
,
J.
, and
Coauthors
,
2001
:
Seasonal predictions based on two dynamical models
.
Atmos.–Ocean
,
39
,
485
501
.
Ding
,
Y.
,
1994
: Monsoon over China. Kluwer Academic, 420 pp.
Gong
,
D.
, and
C.
Ho
,
2003
:
Detection of large-scale climate signals in spring vegetation index (normalized difference vegetation index) over the Northern Hemisphere
.
J. Geophys. Res.
,
108
, 4498, doi:10.1029/2002JD002300.
Gong
,
D.
,
S.
Wang
, and
J.
Zhu
,
2001
:
East Asian winter monsoon and Arctic Oscillation
.
Geophys. Res. Lett.
,
28
(10),
2073
2076
.
Guo
,
D.
, and
Z.
Sun
,
2004
:
Relationships of winter North Pacific Oscillation anomalies with the East Asian winter monsoon and the weather and climate in China (in Chinese)
.
J. Nanjing Inst. Meteor.
,
27
,
461
470
.
Hoerling
,
M.
,
J. W.
Hurrell
, and
T.
Xu
,
2001
:
Tropical origins for recent North Atlantic climate change
.
Science
,
292
,
90
92
.
Hurrell
,
J.
,
1995
:
Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation
.
Science
,
269
,
676
679
.
Hurrell
,
J.
,
1996
:
Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature
.
Geophys. Res. Lett.
,
23
(6),
665
668
.
Jia
,
X.
, and
H.
Lin
,
2011
:
Influence of forced large-scale atmospheric patterns on surface air temperature in China
.
Mon. Wea. Rev.
,
139
,
830
852
.
Jia
,
X.
,
H.
Lin
, and
J.
Derome
,
2009
:
The influence of tropical Pacific forcing on the Arctic Oscillation
.
Climate Dyn.
,
32
,
495
509
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Keenlyside
,
N.
,
M.
Latif
,
M.
Botzet
,
J.
Jungclaus
, and
U.
Schulzweida
,
2005
:
A coupled method for initializing El Niño Southern Oscillation forecasts using sea surface temperature.
Tellus,
57A
,
340
356
.
Kharin
,
V.
,
Q.
Teng
, and
F.
Zwiers
,
2009
:
Skill assessment of seasonal hindcasts from the Canadian historical forecast project
.
Atmos.–Ocean
,
47
, 204–223.
Kumar
,
K.
,
M.
Hoerling
, and
B.
Rajagopalan
,
2005
: Advancing dynamical predication of Indian monsoon rainfall. Geophys. Res. Lett.,32, L08704, doi:10.1029/2004GL021979.
Lee
,
J.-Y.
, and
Coauthors
,
2010
:
How are seasonal prediction skills related to models’ performance on mean state and annual cycle?
Climate Dyn.
,
35
,
267
283
.
Li
,
C.
,
1990
:
Interaction between anomalous winter monsoon in East Asia and El Niño events
.
Adv. Atmos. Sci.
,
7
,
36
46
.
Li
,
S.
,
S.
Hoerling
,
S.
Peng
, and
K. M.
Weickmann
,
2006
:
The annular response to tropical Pacific SST forcing
.
J. Climate
,
19
,
1802
1819
.
Limpasuvan
,
V.
, and
D.
Hartmann
,
1999
:
Eddies and the annular modes of climate variability
.
Geophys. Res. Lett.
,
26
(20),
3133
3136
.
Lin
,
H.
,
J.
Derome
, and
G.
Brunet
,
2005a
: Correction of atmospheric dynamical seasonal forecasts using the leading ocean-forced spatial patterns. Geophys. Res. Lett.,32, L14804,
doi:10.1029/2005GL023060
.
Lin
,
H.
,
J.
Derome
, and
G.
Brunet
,
2005b
: Tropical Pacific link to the two dominant patterns of atmospheric variability. Geophys. Res. Lett.,32, L03801, doi:10.1029/2004GL021495.
Lin
,
H.
,
G.
Brunet
, and
J.
Derome
,
2008
:
Seasonal forecasts of Canadian winter precipitation by post-processing GCM integrations
.
Mon. Wea. Rev.
,
136
,
769
783
.
Melia
,
D. S.
,
2002
:
A global coupled sea ice–ocean model
.
Ocean Modell.
,
4
,
137
172
.
Mitchell
,
T. D.
, and
P. D.
Jones
,
2005
:
An improved method of constructing a database of monthly climate observations and associated high-resolution grids
.
Int. J. Climatol.
,
25
,
693
712
.
Pozo-Vazquez
,
D.
,
M. J.
Esteban-Parra
,
F. S.
Rodrigo
, and
Y.
Castro-Diez
,
2001
:
The association between ENSO and winter atmospheric circulation and temperature in the North Atlantic region
.
J. Climate
,
14
,
3408
3420
.
Rayner
,
N.
,
D. E.
Parker
,
E. B.
Horton
,
C. K.
Folland
,
L. V.
Alexander
,
D. P.
Rowell
,
E. C.
Kent
, and
A.
Kaplan
,
2003
:
Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century
.
J. Geophys. Res.
, 108, 4407,
doi:10.1029/2002JD002670
.
Richman
,
M.
,
1986
:
Rotation of principal components
.
Int. J. Climatol.
,
6
,
293
335
.
Rogers
,
J.
,
1990
:
Patterns of low frequency monthly sea level pressure variability (1899–1996) and associated wave cyclone frequencies
.
J. Climate
,
3
,
1364
1379
.
Shukla
,
J.
, and
Coauthors
,
2000
:
Dynamical seasonal prediction
.
Bull. Amer. Meteor. Soc.
,
81
,
2593
2606
.
Solomon
,
A.
, and
M.
Newman
,
2012
:
Reconciling disparate twentieth-century Indo-Pacific Ocean temperature trends in the instrumental record
.
Nat. Climate Change
,
2
,
691
699
.
Thompson
,
D.
, and
J.
Wallace
,
1998
:
The Arctic Oscillation signature in wintertime geopotential height and temperature fields
.
Geophys. Res. Lett.
,
25
(9),
1297
1300
.
Trenberth
,
K. E.
,
G. W.
Branstator
,
D.
Karoly
,
A.
Kumar
,
N.
Lau
, and
C.
Ropelewski
,
1998
: Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperature. J. Geophys. Res.,103 (C7), 14 291–14 324.
Wallace
,
J.
, and
D. S.
Gutzler
,
1981
:
Teleconnections in the geopotential height field during the Northern Hemisphere winter
.
Mon. Wea. Rev.
,
109
,
784
812
.
Walter
,
K.
, and
H.
Graf
,
2002
: On the changing nature of the regional connection between the North Atlantic Oscillation and sea surface temperature. J. Geophys. Res.,107, 4338, doi:10.1029/2001JD000850.
Wang
,
B.
,
R.
Wu
, and
X.
Fu
,
2000
:
Pacific–East Asian teleconnection: How does ENSO affect East Asian climate?
J. Climate
,
13
,
1517
1536
.
Wang
,
B.
,
R.
Wu
, and
T.
Li
,
2003
:
Atmosphere–warm ocean interaction and its impact on Asian–Australian monsoon variation
.
J. Climate
,
16
,
1195
1211
.
Wang
,
B.
, and
Coauthors
,
2009
:
Advance and prospectus of seasonal prediction: Assessment of APCC/CLiPAS 14-model ensemble retrospective seasonal prediction (1980–2004)
.
Climate Dyn.
,
33
,
93
117
.
Wang
,
L.
,
W.
Chen
, and
R.
Huang
,
2007
: Changes in the variability of North Pacific Oscillation around 1975/1976 and its relationship with East Asian winter climate. J. Geophys. Res.,112, D11110,
doi:10.1029/2006JD008054
.
Watanabe
,
M.
,
2004
:
Asian jet waveguide and a downstream extension of the North Atlantic Oscillation
.
J. Climate
,
17
,
4674
4691
.
Weisheimer
,
A.
, and
Coauthors
,
2009
: ENSEMBLES: A new multi-model ensemble for seasonal-to-annual predictions—Skill and progress beyond DEMETER in forecasting tropical Pacific SSTs. Geophys. Res. Lett.,36, L21711, doi:10.1029/2009GL040896.
Wu
,
A.
, and
W.
Hsieh
,
2004
:
The nonlinear association between ENSO and the Euro-Atlantic winter sea level pressure
.
Climate Dyn.
,
23
,
859
868
.
Wu
,
R.
, and
B.
Kirtman
,
2005
:
Roles of Indian and Pacific Ocean air–sea coupling in tropical atmospheric variability
.
Climate Dyn.
,
25
,
155
170
.
Wu
,
R.
,
Z.
Hu
, and
B.
Kirtman
,
2003
:
Evolution of ENSO-related rainfall anomalies in East Asia
.
J. Climate
,
16
,
3742
3758
.
Yang
,
S.
,
K.-M.
Lau
, and
K.-M.
Kim
,
2002
:
Variation of the East Asian jet stream and Asian–Pacific–American winter climate anomalies
.
J. Climate
,
15
,
306
325
.
Zhang
,
R.
,
A.
Sumi
, and
M.
Kimoto
,
1996
:
Impact of El Niño on the East Asian monsoon: A diagnostic study of the 86/87 and 91/92 events
.
J. Meteor. Soc. Japan
,
74
,
49
62
.
Zhong
,
A.
,
H.
Hendon
, and
O.
Alves
,
2005
:
Indian Ocean variability and its association with ENSO in a global coupled model
.
J. Climate
,
18
,
3634
3649
.