HTML
-
A large-scale extratropical teleconnection pattern in boreal summer, named the Asian-Pacific Oscillation (APO), has recently been identified. It is characterized by a zonal seesaw of midlatitudinal upper-tropospheric temperature between Asia and the Pacific. The high (low) tropospheric eddy temperature over Eurasia is usually accompanied by low (high) eddy temperature over the North Pacific (Zhao et al., 2007). The APO exhibits notable interannual and interdecadal variability, and appears in the non-summer seasons as well (Zhao et al., 2008).
Previous studies have revealed that the APO is closely linked to the Asian summer monsoon and monsoonal precipitation (Zhao et al., 2007). Corresponding to a higher APO index, low-level anomalous southerly wind prevails over the midlatitudes of East Asia, and anomalous westerly wind occupies South Asia and the South China Sea region, which indicates a strengthened Asian summer monsoon. As a result, precipitation increases over South Asia and the northern and southern sides of the Yangtze River, but decreases around the Yangtze River and the Philippines. Therefore, the APO index can be used to indicate the variability of the Asian monsoon and rainfall (Zhao et al., 2007, 2008). From a decade with low APO to one with high APO, rainfall generally enhances over the Asian monsoon region and decreases over North America (Zhao et al., 2011). In addition, the relationship between the APO index and Asian monsoonal precipitation has also been investigated on the interdecadal time scale (Zhou et al., 2009; Liu et al., 2011).
The APO is positively correlated with tropical cyclone frequency in the western North Pacific. When the APO is above (below) normal in summer, more (fewer) tropical cyclones tend to form in the western North Pacific (Zhou et al., 2008; Zou and Zhao, 2011). The APO index has a significant negative correlation with the western Pacific subtropical high (Huang et al., 2013). Associated with the variation of the summer APO, significant anomalous circulation signals can even be observed over the Asia-Pacific-America sector (Zhou and Zhao, 2010). Furthermore, the APO's variability is closely linked with SST in the Pacific, with a significant positive (negative) correlation between the APO index and SST over the extratropical North Pacific (tropical eastern Pacific) on the interannual timescale (Zhou et al., 2010; Zhao et al., 2010).
Therefore, exploring the physical mechanism responsible for the formation and maintenance of the APO is necessary for predicting the variation in the climate of the Northern Hemisphere. (Zhao et al., 2008) suggested that the formation of the APO is related to the difference in solar radiation between the Asian continent and the North Pacific. The thermal effect of the Tibetan Plateau (TP) intensifies the temperature of the local troposphere and decreases the tropospheric temperature over the North Pacific through zonal and vertical circulations, leading to the formation of the APO. A number of attempts have been made to understand the mechanism responsible for the formation of the APO and its associated climate anomalies by using global climate models. As a result, it has been proven that the major characteristics and dynamical structures of the APO in summer can be captured by some coupled climate system models (CSMs; Zhao et al., 2010; Man and Zhou, 2011; Chen et al., 2013b), and the relationship between the APO and Pacific SST can be successfully reflected in CCSM3 simulations (Nan et al., 2009; Zhao et al., 2010). Recently, (Huang et al., 2013) assessed the predictability of the summer APO index using the European Multi-model Ensemble System and found that these models can predict the interannual variability of the summer APO well. (Chen et al., 2013a) further indicated that the summer APO and its associated climate anomalies can be predicted by NCEP CFSv2 by up to 5 months in advance.
The above studies illustrate the importance of evaluating the capability of various climate models in simulating the characteristics of the APO when attempting to predict circulation anomalies associated with it. The capabilities of CSMs to realistically reproduce the current state of regional and global climate is vitally important for reliable projections of climate change in the future (Kidston and Gerber, 2010). However, few efforts have been made to investigate the impact of model resolution on the simulation of the APO. Since the complexity of the topography and underlying surface state can be better described in climate models with relatively higher resolution, high-resolution model simulations provide us with an opportunity to analyze the subsequent influence of such high horizontal resolutions on the simulation of the APO.
The present study evaluates the ability of two versions of BCC-CSM, with different resolutions, i.e., BCC_CSM1.1 and BCC_CSM1.1(m), in simulating the variability of the large-scale APO pattern and associated atmospheric circulation anomalies. The aim is to address whether the observed characteristics of the summer APO can be reproduced in the two models, and, if so, what the impact is of the higher horizontal resolution in BCC_CSM1.1(m) on the simulation of the characteristics of the summer APO. Reasons for any identified improvements in the simulation of the summer APO by BCC_CSM1.1(m) will then be identified.
Following this introduction, the models, datasets and method applied in the study are described in section 2. Section 3 presents the observed characteristics of the APO and associated precipitation, as well as the model results. The possible reasons for any identified improvements in the simulation of the APO by BCC_CSM1.1(m) are discussed in section 4. Finally, conclusions and a discussion are provided in section 5.
-
BCC_CSM is a coupled climate system model, including atmosphere, ocean, land surface and sea ice components. There are two versions of the model system with different horizontal resolutions, i.e., BCC_CMS1.1 and BCC_ CSM1.1(m). Both models have been involved in CMIP5 (Jiang et al., 2012). A comprehensive atmospheric general circulation model (BCC_AGCM2.1) derived from NCAR CAM3 and modified by (Wu et al., 2008), version 4 of the GFDL's MOM (Griffies et al., 2004), GFDL's Sea Ice Simulator (Winton, 2000), and version 1.0 of the BCC's Atmosphere and Vegetation Interaction Model (Ji et al., 2008), are interactively coupled in BCC_CSM1.1 using version 5 of the NCAR's coupler (Wu, 2012). No flux adjustment is implemented in BCC_CSM1.1. The OGCM, MOM, has 40 vertical layers and the nominal horizontal resolution is 1° × 1°, with equatorial refinement to 0.33° between 30°S and 30°N. The horizontal resolution and the sea-land distribution in the sea ice model are the same as that in MOM. BCC_CSM1.1(m) is an advanced version of BCC_CSM1.1, with a moderate atmospheric resolution. Compared to BCC_AGCM2.1, which is used in BCC_CSM1.1 and runs at a T42 spectral resolution (approximately 2.8°× 2.8°), the atmospheric component of BCC_CSM1.1(m) is BCC_AGCM2.2, which runs at a T106 horizontal resolution (approximately 1.125°× 1.125°). Both models use a terrain-following vertical hybrid sigma-pressure coordinate, with 26 levels and a rigid lid at 2.914 hPa. The dynamical framework and physical processes are the same in the two models and are fully introduced in Wu et al. (2008, 2010). A fair number of studies have evaluated the performance of the two models in climate simulation and the projection of future climate change, especially regarding the simulation of precipitation and temperature fields (Gao et al., 2012; Gao et al., 2013; Xin et al., 2013). Nevertheless, many previous studies have focused mainly on changes in surface air temperature. Few researchers have paid attention to the models' abilities in reproducing the variation in upper-tropospheric eddy temperature (Zhou and Zhang, 2012).
Monthly model outputs of the CMIP5 historical simulation experiments of the two models for 31 summers (June-July-August) from 1979 to 2009 are used in this study. The external forcing of the historical simulations changes with time, including mixed greenhouse gases (CO2, N2O, CH4, CFC11, and CFC12), aerosols, ozone, volcanoes and solar radiation. All the forcing data are provided by CMIP5, except for volcanoes. The temporal resolution of CO2 emissions and solar radiation is 1 year, while the time interval of the aerosol data is 10 years. For the purpose of comparison with other model results, the present study uses the observational data from NCEP-DOE Reanalysis-2 (Kanamitsu et al., 2002) and CMAP (Xie and Arkin, 1997) to validate the simulations and discuss the biases. The data after 1979 are chosen because satellite observations become available since then and the reanalysis data are more reliable and homogeneous than during the pre-satellite period (Dell'Aquila et al., 2005).
EOF analysis is applied to the eddy temperature to identify the APO teleconnection pattern over the Northern Hemisphere. A weighting by latitude is applied to the EOF results. Regression and correlation analyses are conducted to explore relationships between pairs of variables, while the statistical significance of correlation coefficients, regression values and long-term variation trends are assessed using the Student's t-test. Besides, in order to compare the simulations with observations, the model results are interpolated to the observational grids using the bilinear interpolation method.
-
Since the APO is defined by the upper-tropospheric (500-200 hPa) eddy temperature (T'; Zhao et al., 2007), which is obtained by removing the zonal mean temperature \((\overline{T})\) from the total air temperature (T), i.e., \(T'=T-\overline{T}\), we first investigate the distributions of observed and simulated summer mean total air temperature and eddy temperature over the upper troposphere (Fig. 1). The observed T gradually decreases from the south to the north, and the main temperature band is oriented from west to east, with a maximum temperature center greater than -20°C located over the southern flank of the TP (Fig. 1a). Compared with the observation, both models capture the distributional feature of T, which decreases from low to high latitudes. However, BCC_CSM1.1 systematically underestimates the intensity of T. The range of the simulated main temperature band is much smaller than the observed result, and the temperature center is mainly located in the TP region (Fig. 1b). Compared with BCC_CSM1.1, BCC_CSM1.1(m) realistically simulates not only the temperature band that is oriented from west to east, but also the maximum temperature center. This result indicates a remarkable improvement in reproducing the upper-tropospheric total air temperature by BCC_CSM1.1(m), due to its higher horizontal resolution (Fig. 1c).
The features of observed and simulated eddy temperature are further examined in Fig. 1. A prominent out-of-phase variational pattern of eddy temperature exists in the midlatitudes between Asia and the Pacific in boreal summer. Positive values occupy the lower and middle latitudes of Asia, with a high temperature center of 4°C located over the TP region. Negative values appear over the central-eastern Pacific, with a minimum value of -3°C (Fig. 1a). Additionally, another center of negative eddy temperature occurs over the Atlantic region. These features are realistically captured by the two models, except that the value of the positive eddy temperature center simulated by BCC_CSM1.1 is about 6°C, which is approximately 2°C higher than observed; plus, the simulated negative eddy temperature center is about -2°C, which is lower than observed by approximately 1°C. Compared with the simulation of BCC_CSM1.1, the simulated positive center over the TP region by BCC_CSM_1.1(m) is closer to the observation, but the simulated negative center over the central-eastern Pacific is located further eastward than observed.
Figure 1. The climatology of the summer mean upper-tropospheric (500-200 hPa) $T$ (color-shaded; units: $^\circ$C) and $T'$ (contours; units: $^\circ$C) during 1979-2009 for (a) NCEP, (b) BCC_CSM1.1, and (c) BCC_CSM1.1(m), respectively. The thick black line denotes the orographic isocline of the main body of the TP. And the green boxes represent the key regions selected to construct the APO index.
Figure 2. Differences in climatological $T$ (color-shaded; units: $^\circ$C) and $T'$ (contours; units: $^\circ$C) between (a) BCC_CSM1.1 and NCEP, (b) BCC_CSM1.1(m) and NCEP, and (c) BCC_CSM1.1(m) and BCC_CSM1.1. The thick green line indicates the orographic isocline of the main body of the TP. The thick green line indicates the topographic contour of 3000 m.
The difference between the model results and the reanalysis data shows that the climatological total air temperature is notably underestimated throughout the Northern Hemisphere in the results of BCC_CSM1.1. The simulated warm biases of eddy temperature appear along the northern flank of the TP, while cold biases mainly occur over South America (Fig. 2a). Compared with that in BCC_CSM1.1, the intensity of simulated total air temperature is effectively enhanced in BCC_CSM1.1(m) (Fig. 2c), especially over midlatitude areas. Warm biases of the simulated total air temperature in BCC_CSM1.1(m) mainly occur over the central-western Pacific, while cold biases mainly appear over South America. The distribution of the simulated eddy temperature bias in BCC_CSM_1.1(m) is similar to that of the total air temperature. The positive anomaly center is situated over the central Pacific, with a maximum value exceeding 2.5°C; and the negative anomaly center lies over South America, with a minimum value lower than -2°C (Fig. 2b).
The Taylor diagram between outputs of the two models and observations shown in Fig. 3 further illustrates that, although BCC_CSM1.1(m) is better able to reproduce the upper-tropospheric total air temperature than BCC_CSM1.1, the eddy temperature simulated by BCC_CSM1.1 is overall more agreeable with the observation, not only in the Northern Hemisphere but also in regions over East Asia and the North Pacific. Meanwhile, it is noteworthy that both models exhibit better capacity for simulating the eddy temperature over East Asia than over the North Pacific, which is contrary to the results of previous model studies (Huang et al., 2013; Chen et al., 2013a). Possible reasons will be discussed in section 5.
Figure 3. Taylor diagram comparing the spatial statistics between the simulations of the two BCC models and observations for the summer mean $T'$ over the Northern hemisphere (NH), East Asia (EA) and the North Pacific (NP) regions during 1979-2009. "REF" denotes the NCEP-DOE reanalysis; the azimuth angle indicates the spatial correlation coefficient between observations and model outputs; the radial distance indicates the standard deviation between observations and model outputs; and the distance from "REF" represents the centralized RMSE.
-
Because the APO is identified through the contrast in upper-tropospheric eddy temperature between Asia and the North Pacific, one may speculate that the reasonable simulation of climatological eddy temperature in BCC_CSM1.1 should lead to a better simulation of the characteristics of the APO. To examine this assertion, the simulated spatial pattern and interannual variability of the APO are examined in this subsection.
Following Zhao et al. (2010, 2012), an EOF analysis of the observed and simulated eddy temperature is performed for the period 1979-2009 to reveal the teleconnection pattern over the Northern Hemisphere in summer. The regression map of the vertically integrated eddy temperature from observations and simulations between 500 and 200 hPa with respect to the normalized PC1 is shown in Fig. 4. The first EOF mode of observations accounts for 20.3% of the total variance, manifesting a prominent out-of-phase relationship of temperature between Africa-Eurasia and the North Pacific region. Two positive centers are located over North Africa and East Asia, respectively, with the maximum value exceeding 0.04, while a negative center is situated over the North Pacific, with a minimum value below -0.04 (Fig. 4a). Compared with observations, it is found that BCC_CSM1.1 can reproduce the negative center over the North Pacific, but fails to reproduce the two positive centers over North Africa and East Asia. Instead, it produces a false positive center over South America (Fig. 4b). The EOF1 pattern simulated by BCC_CSM1.1(m) agrees well with the observation, and the out-of-phase relationship of temperature between Africa-Eurasia and the North Pacific region in the Northern Hemisphere is reflected well. The spatial correlation coefficients between observations and outputs of BCC_CSM1.1 and BCC_CSM1.1(m) are 0.40 and 0.77, respectively, indicating that the simulated spatial pattern of the APO is greatly improved in BCC_CSM1.1(m).
Figure 4. Observed and simulated regression maps of the vertically integrated summer mean eddy temperature between 500 and 200 hPa with respect to the normalized PC1 over the Northern Hemisphere during 1979-2009, based on (a) NCEP-DOE, (b) BCC_CSM1.1, and (c) BCC_CSM1.1(m).
Figure 5. Normalized summer (a-c) AI, (d-f) PI, and (g-i) APO indices during 1979-2009, based on NCEP-DOE (upper panels), BCC_CSM1.1 (middle panels), and BCC_CSM1.1(m) (lower panels). The detrended correlation coefficient (CC-I) between observations and model results is marked in the top-right corner of each panel, and the solid line indicates the linear trend of the index.
To investigate the interannual variability of the APO, we define the APO index as the arithmetic difference between the Asian tropospheric eddy temperature index (AI) and the North Pacific tropospheric eddy temperature index (PI), where the AI and PI are computed by the regionally averaged upper-tropospheric (500-200 hPa) T' over (15°-45°N, 70°-110°E) and (15°-45°N, 170°-110°W), respectively (Zhao et al., 2007; Huang et al., 2013). The above selected key regions for AI and PI are marked in Fig. 1. The capability of the models in reproducing the APO's interannual variability is measured by the detrended correlation coefficients (CC-I) between the observed and simulated AI, PI and APO indices. Figure 5g shows that the observed APO index exhibits significant interannual variability in the last 30 years, with a linear descending trend of -0.041°C yr-1 (exceeding the 95% confidence level), which is consistent with previous findings (Huang et al., 2013). The variational trend of the AI is consistent with that of the APO index (Fig. 5g), while the PI displays a weak ascending trend that is not significant (Fig. 5d). Therefore, the weakening trend of the APO index can be mainly attributed to the decreasing trend of the upper-tropospheric eddy temperature over land, implying an enhanced thermal contrast between Asia and the North Pacific in recent decades. Since the APO index has a distinct linear trend, all the linear trends of the indices are removed to investigate the interannual variability of the APO.
Figure 6. Regression maps of summer precipitation (units: mm d$^-1$) with reference to the normalized APO index for (a) CMAP, (b) BCC_CSM1.1, and (c) BCC_CSM1.1(m). The blue (brown) color indicates the negative (positive) precipitation anomalies. Areas of light (dark) shading are values at/above the 90% (95%) confidence level. The thick green line indicates the topographic contour of 3000 m.
The CC-I values of the AI, PI and APO index are -0.17, 0.02 and -0.09, respectively, in BCC_CSM1.1, which are very low and not significant. Compared with the observation, BCC_CSM1.1 fails to reproduce the interannual variability of the APO index. Further analysis indicates that this failure can be attributed to the unreasonable simulation of the AI and PI, suggesting that BCC_CSM1.1 performs poorly in reproducing the variation of the APO. Although a linear trend can be detected in BCC_CSM1.1, the result is opposite to that derived from the observation (Fig. 5; middle panels). Figure 5 (lower panels) displays the evolution of the summer AI, PI and APO index from 1979 to 2009 in BCC_CSM1.1(m). The CC-I values of the AI, PI and APO index are 0.35, 0.33 and 0.40, respectively, and all exceed the 95% confidence level. These results demonstrate that BCC_CSM1.1(m) is more capable than BCC_CSM1.1 when it comes to reproducing the interannual variability of the APO. However, the linear decreasing trend of the APO index is not significant in the results of BCC_CSM1.1(m), suggesting that it lacks skill in reproducing the long-term variability of the APO pattern. Note that this phenomenon is also found in some other climate system models (Huang et al., 2013).
-
Figure 6 displays regression maps of precipitation against the APO index from the model outputs and observations. The observed result (Fig. 6a) shows that, corresponding to the APO's positive phase, positive precipitation anomalies mainly occur over the summer monsoon regions of the Northern Hemisphere, including Mexico, East Asia, South Asia and West Africa. Positive precipitation anomalies also occur in the lower-latitude region of the central Pacific. On the contrary, negative precipitation anomalies appear over the western and northern parts of the monsoon regions, such as extratropical North America, central-western Asia, the Middle East, and North Africa, where the monsoon-desert coupling phenomenon is distinct (Zhao et al., 2007). The above results derived from observations are consistent with previous findings (Zhao et al., 2012).
Figure 6c presents a regression map of summer precipitation with respect to the APO index in BCC_CSM1.1(m). In general, BCC_CSM1.1(m) captures the characteristics of the rainfall distribution associated with the APO index well, such as the positive precipitation anomaly band extending from 30°E to 150°W, as well as the negative precipitation anomaly band along 40°N that extends from western Africa to the extratropical Pacific and North America. Positive precipitation anomalies are mainly found over the tropics of South America, Mexico's monsoon region, the subtropics of central-western Pacific, the Indochina Peninsula, India, and tropical North Africa. However, the simulated positive precipitation anomalies in Mexico and tropical North Africa are less significant than observed. The simulated relationship between precipitation and the APO index in BCC_CSM1.1(m) agrees qualitatively with the observation.
Figure 7. Regression maps of (a-c) 200 hPa (left panels) and (d-f) 850 hPa (right panels) winds (units: m s$^-1$) against the normalized APO index in summer for (a, d) NCEP-DOE reanalysis, (b, e) BCC_CSM1.1, and (c, f) BCC_CSM1.1(m), respectively. Shaded areas are values exceeding the 95% confidence level; the prevailing wind is indicated by the red arrow, and the "C" ("A") denotes the anomalous cyclone (anticyclone) center. The thick dashed line denotes the orographic isoline of 1500 m in right panels.
Compared with the observation, BCC_CSM1.1 fails to reproduce the relationship between the simulated precipitation anomalies and the APO index (Fig. 6b). The positive precipitation anomalies over the summer monsoon regions of the Northern Hemisphere are largely underestimated, especially over India, the Indochina Peninsula, and Mexico. The simulated negative rainfall anomaly band along 40°N is also not consistent with observations, especially over central-western Asia and North America. Moreover, the simulated precipitation anomalies over the Pacific are much less significant than observed, indicating a lower ability of BCC_CSM1.1 in reproducing the APO-related precipitation in the Northern Hemisphere, as compared with BCC_CSM1.1(m).
3.1. Simulation of upper-tropospheric temperature
3.2. Simulation of the APO
3.3. Simulation of APO-related precipitation
-
Previous studies have revealed that the APO's variability is closely linked to large-scale atmospheric circulation anomalies that directly influence APO-related precipitation, such as the extratropical westerly jet over Eurasia, the western Pacific subtropical high, the South Asian high, and the Asian summer monsoon (Zhao et al., 2010, 2012; Huang et al., 2013). Thus, we further examine the APO-related atmospheric circulation anomalies simulated by the models.
Figure 7 separately shows regression maps of winds at 200 hPa and 850 hPa with respect to the summer APO index from 1979 to 2009, based on observations and the simulations of the two models. As shown in Fig. 7a, a positive APO index is associated with the large-scale anomalous anticyclonic circulation that covers the midlatitude region from 60°E to 150°W at 200 hPa, with three centers near western Asia, northeastern Asia and the central-northern Pacific, respectively. This anticyclonic anomaly actually corresponds to the northward-extended and intensified South Asian high. Meanwhile, anomalous easterly winds prevail from the central-eastern Pacific to Eurasian regions, and westerly anomalies appear along the northern flank of the aforementioned anomalous anticyclone centers over Eurasia and the North Pacific, indicating an enhanced extratropical upper-level westerly jet stream and a strengthened summer monsoon over South Asia and the South China Sea. The observed anomalous atmospheric circulations in the upper troposphere related to the APO are consistent with previous findings (Zhao et al., 2007; Huang et al., 2013). Figure 7b shows that BCC_CSM1.1 fails to reproduce the anomalous anticyclone center over Eurasia, and the simulated easterly wind anomalies are much weaker and less significant than observed, especially over the North Pacific. Compared with BCC_CSM1.1, BCC_CSM1.1(m) generally reproduces the circulation anomalies associated with the APO; namely, the strengthened and northward-extended South Asian high, the intensified extratropical westerly jet, and the tropical easterly jet. This result suggests that the variation of the Asian summer monsoon associated with a positive APO index can be reasonably reproduced in BCC_CSM1.1(m). However, the simulated anomalous anticyclone center over North Pacific is located further eastward than observed (Fig. 7c). Figure 7d further shows that when the APO index is above normal, two anomalous anticyclone centers emerge in the central-northern Pacific and Japan, respectively, at 850 hPa. Easterly wind anomalies extend from the central Pacific to southern Japan, and anomalous southerly winds are observed over northeastern China, indicating that the East Asian summer monsoon intensifies. Meanwhile, westerly wind anomalies prevail from the western Indian Ocean to the South China Sea, corresponding to the strengthened southwesterly monsoonal flow over these areas when the APO index is positive. The above atmospheric circulation anomalies in the lower troposphere are similar to results published in earlier studies (Zhao et al., 2007, 2012). Compared with the observation, the anomalous anticyclone center situated over Japan is not reproduced in the simulation of BCC_CSM1.1, and the simulated easterly wind anomalies over the central Pacific and the southwesterly wind anomalies over the Indian Ocean are not as significant as they are in the observations, implying an underestimation of the anomalous circulations in the lower troposphere associated with the variation of the Asian summer monsoon (Fig. 7e). Figure 7f illustrates that, not only the anomalous anticyclone centers, but also the prevailing wind anomalies associated with the APO index, are well represented in BCC_CSM1.1(m). Apparently, BCC_CSM1.1(m) has significantly improved the simulation of the atmospheric circulation anomalies over the lower troposphere, as compared with BCC_CSM1.1.
Overall, the variational features of the Asian summer monsoon are better captured by BCC_CSM1.1(m) than BCC_CSM1.1. This is the reason why the simulation of monsoonal precipitation is improved in BCC_CSM1.1(m), as shown in Fig. 6. Precipitation anomalies around the TP can induce soil moisture variation, and lead to increases in the tropospheric temperature and intensification of the APO pattern (Liu et al., 2015). Therefore, compared with BCC_CSM1.1, the realistic simulation of the Asian summer monsoon's variation in BCC_CSM1.1(m) could be a possible contributor to reproducing a more reasonable APO, since the interannual variability of the APO is closely linked to the variation of the Asian summer monsoon and monsoonal precipitation (Zhao et al., 2007)
-
The above analysis demonstrates that BCC_CSM1.1(m) exhibits an encouraging ability to reproduce not only the interannual variation of the APO, but also the precipitation and atmospheric circulation anomalies related to the APO. Compared to BCC_CSM1.1, BCC_CSM1.1(m) has remarkably improved the simulation of the characteristics of the APO in summer. To better understand the reason why the finer horizontal resolution of BCC_CSM1.1(m) leads to a more reasonable simulation of the APO, we further investigate the observed and simulated APO-related surface temperature, which plays a crucial role in the formation and maintenance of the APO (Zhao et al., 2008). Figure 8 presents regression maps of surface temperature against the APO index in summer, from observations and model results. Both the surface air temperature over land and SST are used for computation.
As shown in Fig. 8a, the APO's variability is closely linked to SST anomalies over the Pacific. When the APO index is above normal, large-scale significant and positive SST anomalies appear over the extratropical North Pacific from 120° E to 180°E. Meanwhile, significantly negative SST anomalies prevail over the tropical eastern Pacific. Such a spatial pattern of SST anomalies suggests a potential linkage between the APO and ENSO, which is consistent with the findings of (Zhou et al., 2010). However, sensitivity experiments have demonstrated that the SST anomalies over the two regions have opposite impacts on the intensity of the APO (Zhao et al., 2010), and the simultaneous SST variation over the tropical eastern Pacific alone cannot trigger a large-scale teleconnection pattern like the APO (Zhao et al., 2008). Thus, the tropical eastern Pacific SST anomalies captured by both models are not the key factor in determining the APO characteristics in summer. Figure 8b shows that, although BCC_CSM1.1 can reproduce the negative SST anomalies over the tropical eastern Pacific, it fails to capture the positive SST anomalies in the midlatitudes of the North Pacific. (Zhou et al., 2009) proposed that a significantly positive correlation exists between the APO and North Pacific SST on the interannual timescale. When the APO index is above normal, an anomalous anticyclone dominates the lower troposphere over the North Pacific, which is favorable for a warming of SST in the North Pacific. Meanwhile, negative heat fluxes appear in the North Pacific, accompanied by an intensification of northward warm water advection. All these factors are favorable for a warming of SST in the North Pacific. Compared with that in BCC_CSM1.1 (Fig. 8b), the regression map of positive SST anomalies over the North Pacific in BCC_CSM1.1(m) (Fig. 8c) is more consistent with the observation (Fig. 8a), which contributes to the improvement in the BCC_CSM1.1(m) simulation of the characteristics of the APO. This may also explain why the simulated PI is more realistic in BCC_CSM1.1(m) than in BCC_CSM1.1, since the North Pacific is a key region for defining the PI (Fig. 5).
Figure 8. Regression maps of surface air temperature (units: $^\circ$C) over land and SST (units: $^\circ$C) over ocean, with respect to the summer APO index, based on (a) NCEP-DOE reanalysis, (b) BCC_CSM1.1, and (c) BCC_CSM1.1(m). Shaded areas are values exceeding the 95% confidence level. The thick green line indicates the topographic contour of 3000 m.
On the other hand, the formation of the APO is closely correlated with an elevation in the heating effect of the TP (Zhao et al., 2008). In a recent study, (Liu et al., 2015) proposed a new physical mechanism linking winter Pacific SST to the subsequent summer's APO. They pointed out that the previous winter's Pacific SST anomalies can persist until spring to cause an SLP anomaly over the North Indian Ocean in the subsequent spring and summer. The latter induces anomalous vertical motion that modulates the surface air temperature over the southern and western TP, maintaining the summer APO. The regression map of observed surface temperature shows that, associated with a positive APO index, prominent negative surface air temperature anomalies appear over the southern TP, while positive surface air temperature anomalies occur over the western TP region. Meanwhile, negative SST anomalies emerge in the northern Indian Ocean (Fig. 8a). The observed variation in surface temperature related to the APO index is highly consistent with the results of Liu et al. (2015, Fig. 5a). Compared with the observation, the variational features of the surface temperature associated with the APO are successfully reproduced in BCC_CSM1.1(m), i.e., both the negative (positive) surface air temperature anomalies in the southern (western) TP, and the SST anomalies over the northern Indian Ocean associated with the thermal effect of the TP, are successfully captured by BCC_CSM1.1(m) (Fig. 8c). In contrast, the surface air temperature anomalies over the southern and western TP, and the SST anomalies over the northern Indian Ocean, are not reproduced well by BCC_CSM1.1 (Fig. 8b). This result indicates that the physical processes responsible for maintaining the summer APO can be realistically reflected in BCC_CSM1.1(m) but not in BCC_CSM1.1, which explains why BCC_CSM1.1(m) can produce a more reasonable AI (Fig. 5) and is more capable of reproducing the characteristics of the summer APO compared with BCC_CSM1.1.
4.1. Simulation of APO-related atmospheric circulation anomalies
4.2. Simulation of APO-related surface temperature
-
The APO is an important teleconnection pattern that is closely associated with climate variations over the subtropics in summer, especially the Asian-Pacific sector. The APO index is a useful index for assessing large-scale circulation anomalies (Zhao et al., 2007, 2012). In this paper, we use the monthly historical simulations of two versions of BCC_CSM, with different horizontal resolutions, i.e., BCC_CSM1.1 and BCC_CSM1.1(m), along with NCEP-DOE reanalysis data, to evaluate the models' performances in reproducing the interannual variability of the APO. APO-related precipitation and associated atmospheric circulation anomalies are also investigated. The reason for the better simulation of the APO by BCC_CSM1.1(m) is examined. The main results can be summarized as follows:
(1) Both models can capture the spatial distribution of the upper-tropospheric total air temperature, which decreases from low to high latitudes in summer. Compared with BCC_CSM1.1, BCC_CSM1.1(m) effectively increases the simulated intensity of the total air temperature, and the results are closer to observations. However, analysis of the Taylor diagram shows that the simulated eddy temperature in BCC_CSM1.1 is more consistent with the observation, not only over East Asia and the North Pacific, but also over the entire Northern Hemisphere, as compared to that simulated by BCC_CSM1.1(m).
(2) Compared with BCC_CSM1.1, the spatial pattern of EOF1 simulated by BCC_CSM1.1(m) is highly consistent with that from the observation. The spatial correlation coefficients between observations and the outputs of BCC_CSM1.1 and BCC_CSM1.1(m) are 0.40 and 0.77, respectively. Meanwhile, the correlation coefficients between the observed and simulated AI, PI and APO index in BCC_CSM1.1(m) are 0.35, 0.33 and 0.40, respectively, which are much higher than those between observations and the simulations of BCC_CSM1.1. BCC_CSM1.1(m) shows an encouraging capacity to reproduce not only the spatial pattern of the APO, but also the APO's interannual variability, due to its higher horizontal resolution. In particular, BCC_CSM1.1(m) exhibits greater skill in simulating the interannual variability of the eddy temperature index in Asia than it does over the North Pacific.
(3) Based on comparisons between model results and observations, it is found that BCC_CSM1.1(m) can successfully reproduce the APO-related atmospheric circulation anomalies, such as the northward-shifted and intensified South Asian high, the strengthened extratropical westerly jet, and the tropical easterly jet in the upper troposphere, as well as the southwesterly monsoonal flow over the Indian Ocean and the intensified subtropical anticyclone over the North Pacific and Japan in the lower troposphere. As a result, the increased precipitation over tropical North Africa, South Asia and East Asia, and the decreased precipitation over subtropical North Africa, Japan and North America, simulated by BCC_CSM1.1(m), agree qualitatively with observations. In contrast, the circulation anomalies associated with a positive APO index in the simulation of BCC_CSM1.1 are less consistent with observations, which indicates a poor performance of BCC_CSM1.1 in simulating APO-related precipitation.
(4) Regression analysis further indicates that BCC_ CSM1.1(m) can realistically capture SST anomalies over the North Pacific and northern Indian Ocean, as well as the anomalous surface air temperature along the southwestern flank of the TP. These temperature anomalies are closely linked to the maintenance of the APO. However, these relationships are missed by BCC_CSM1.1, suggesting that a higher horizontal resolution is crucial for BCC_CSM to reasonably simulate the physical processes involved in the formation and maintenance of the APO in summer. This may explain why BCC_CSM1.1(m) can reproduce the APO's interannual variability and accompanying circulation anomalies more reasonably than BCC_CSM1.1, and presents a substantial improvement in simulating the characteristics of the APO and APO-related precipitation anomalies.
Although BCC_CSM1.1(m) is capable of simulating the APO teleconnection and its interannual variability in summer, it fails to reproduce the observed long-term variational trend of the APO index and the AI and PI. In fact, this phenomenon is also found in some other CSMs. Since the temporal resolution of aerosols used in most CSMs is 10 years, (Huang et al., 2013) argued that it is hard for models to realistically simulate decadal changes in winter/spring snow depth over the TP under a constant aerosol concentration. This directly influences the simulation of the long-term variation in tropospheric temperature over land areas of Asia.
Besides, several previous studies have revealed that CSMs always demonstrate a higher predictive skill over the North Pacific, as compared to land areas of Asia. The lack of predictability over land areas is possibly associated with the complicated land-atmosphere interaction and feedback processes at play, which may not be represented well in models (Chen et al., 2013a; Huang et al., 2013). Our results support this argument, since the increased horizontal resolution in BCC_CSM1.1(m) helps to improve the description of this complicated land-atmosphere interaction, and subsequently the surface temperature over land (Jiang et al., 2015).
It is worth noting that the results of the present study were obtained based on the historical simulation experiments of the two models. Since the APO is closely linked to the variation of weather and climate, it is also necessary to assess the models' abilities in predicting the APO's variation using results from hindcast experiments. Moreover, the teleconnection pattern over the upper troposphere also exists in other, non-summer, seasons. Further studies that evaluate the performance of the models in simulating and predicting the APO and its associated climate variations in non-summer seasons are needed.