Climates in the recent past offer tangible scenarios by which we may refine our understanding of the Earth system. The mid-Holocene (MH), about 6000 years before the present, interests climatologists, geologists, ecologists and modellers because of its unique climate and a wealth of paleoclimate evidence. Hence, the Paleoclimate Modeling Intercomparison Project (PMIP), which aims to understand the mechanisms of climate change and to test the capability of climate models for representing a climate state different from that of the present day (PD), has focused on the MH climate.
PMIP has already launched two phases of simulation researches on the MH climate. In its initial phase (PMIP1), the MH simulations were performed mainly by using atmospheric general circulation models (AGCMs) (e.g. Hall and Valdes, 1997; Joussaume et al., 1999; Wang, 1999, 2000, 2002). Both the sea-surface temperature and the vegetation over land were kept the same as the present. Thus, these simulations only test the sensitivity of the atmosphere and land surface to the change in insolation caused by the Earth's orbital parameters. Generally, AGCMs could reproduce the enhancement and northward shift of the African and Asian summer monsoons in the MH, but they underestimated the magnitude of monsoon precipitation (Joussaume et al., 1999; Braconnot et al., 2007a). This partially resulted from the drawback of the PMIP1 models in simulating the change of oceanic circulations. Hence, in its second phase (PMIP2), fully coupled ocean–atmosphere general circulation models (OAGCMs) were applied to simulate the MH climate (e.g. Harrison et al., 2002; Liu et al., 2003, 2004; Otto-Bliesner et al., 2006; Braconnot et al., 2007a, 2007b; Ohgaito and Abe-Ouchi, 2007; Marzin and Braconnot, 2009; Dallmeyer et al., 2010). Compared with the PMIP1 simulations, the PMIP2 simulations generally agreed better with proxy data (Braconnot et al., 2007a, Wang et al., 2010), and provided further insights into climate features and associated mechanisms in the MH.
Observations and model studies of the modern climate show that variations of the upper-tropospheric temperature may exert significant impacts on climate anomalies. Zhao et al. (2007b) noticed the out-of-phase pattern named the Asian-Pacific oscillation (APO) in the variability of the upper-tropospheric temperature over the Asian-Pacific region. Namely, high (low) upper-tropospheric temperature over Asia is accompanied with low (high) upper-tropospheric temperature over the North Pacific. For summer, the APO indicates the variability of thermal contrasts between Asia and its adjacent North Pacific and is closely associated with the Northern Hemispheric atmospheric circulations (Zhao et al., 2010), the Asian monsoon precipitation (Zhao et al., 2007b), the tropical cyclone activities over the western North Pacific (Zhou et al., 2008), and the sea-surface temperature in the tropical and North Pacific (Zhao et al., 2010; Zhou et al., 2010). On the interdecadal time scale, the variability of APO is also related to climate anomalies over North America during summer (Zhao et al., 2011). In fact, a stronger APO phenomenon occurred in the summertime of the MH compared to the present climate and affected the atmospheric circulation and monsoon precipitation over Asia (Zhou and Zhao, 2010b).
However, previous works focused on the MH climate and atmospheric circulations in summer and winter while paid little attention to those in spring, especially to the upper-tropospheric temperature change. For the present climate, the East Asian monsoon precipitation begins over southern China in spring, advancing northward to the Yangtze River valley of China and leading to the formation of the Meiyu belt (e.g. Zhao et al., 2007a). Meanwhile, the spring APO is also tightly linked to the atmospheric circulation over the Asian-Pacific region and the East Asian rainfall (Zhou and Zhao, 2010a). Did the APO phenomenon also occur in the springtime of the MH? If yes, how did it link to the regional atmospheric circulation and climate? Was there any difference in the APO change between spring and summer during the MH? With these questions in mind, in the present study, we examine the change of spring APO and associated atmospheric circulations from the CCSM3 output of PMIP2 during the MH.
The rest of the paper is organized as follows. The PMIP2 experimental design and the model data are described in Section 2, and the model ability in simulating upper-tropospheric temperature in spring is evaluated by a comparison with the observation in Section 3. The change of the simulated spring temperature teleconnection in the upper troposphere during the MH is shown in Section 4. In this section, we also present variations of the atmospheric circulation and precipitation in relation to this teleconnection, and compare the CCSM3 results with those of other PMIP2 models. Finally, some concluding remarks and discussions are provided in Section 5.
2. Model and data
In the PMIP2 experimental design (http://pmip2.lsce.ipsl.fr/), the MH simulation was forced by an eccentricity of 0.018682, an obliquity of 24.105°, a precession of 0.87° and an atmospheric CH4 concentration of 650 ppb. The PD simulation was forced by an eccentricity of 0.016724, an obliquity of 23.446°, a precession of 102.04° and an atmospheric CH4 concentration of 760 ppb. Other boundary conditions were the same for both periods. Owing to the difference of orbital parameters, negative solar anomalies occurred in the Northern Hemisphere in the springtime of the MH as compared with the present (Otto-Bliesner et al., 2006). Each simulation was run for several hundred years and reached quasi-equilibrium.
The data used in this study obtained from the PD and MH outputs of the CCSM3 simulation for the last 100 years and are archived in the PMIP2 database. CCSM3 model is a global coupled ocean–atmosphere climate model (Collins et al., 2006). It includes the National Center for Atmospheric Research Community Atmospheric Model version 3 (NCAR-CAM3) AGCM with an approximately horizontal resolution of 2.8° × 2.8° in latitude and longitude and the NCAR-Parallel Ocean Program (POP) oceanic model with a nominal grid spacing of approximately 1° in latitude and longitude.
To validate the PD simulation, the simulated results are compared with the present climatology indicated by the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis data with the resolution of 2.5° × 2.5° in latitude and longitude for 1949–2005 (Kalnay et al., 1996) and the twentieth century reanalysis V2 products with the horizontal resolution of 2° × 2° in latitude and longitude for 1901–2000 (Compo et al., 2006), which may be used to research on the interdecadal variations of large-scale teleconnections over the Northern Hemisphere (Zhao et al., 2011). The statistic significance of the difference between the MH and PD simulations was assessed by the Student's t-test.
3. Comparison between the observed and simulated PD climate characteristic
Figure 1(a) displays the climatological distribution of spring eddy air temperature T′ in the upper troposphere (300–200 hPa) derived from the NCEP/NCAR reanalysis data. Here, T′ = T − [T], where T is the air temperature and [T] is the zonal mean of the air temperature. T′ is a good measure for reflecting the zonal temperature difference, since air temperature decreases from the equator to the pole and exhibits an obvious meridional contrast (Zhao et al., 2007b). As shown in Figure 1(a), positive values appear over Asia with a central value of 3 °C and negative values emerge over the central-eastern Pacific with a central value of − 2 °C. The T′ feature simulated by the PD experiment (Figure 1(b)) is broadly comparable to the observation in both magnitude and spatial distribution. This similarity suggests that the CCSM3 model can well reproduce the observational T′ pattern in the upper troposphere.
An empirical orthogonal function (EOF) analysis was further performed on the spring T′ in the upper troposphere. The first EOF mode (EOF1) from the NCEP/NCAR reanalysis is shown in Figure 2(a), accounting for 25% of the total variance. The positive values over Eurasia and the negative values over the North Pacific can be explicitly observed in this figure, indicating an out-of-phase relationship. To quantitatively measure the upper-tropospheric T′ variation over these two regions, we defined the regional mean upper-tropospheric T′ values over (30–120°E, 5–40°N) and (180–90°W, 5–40°N) respectively as the Asian T′ index (AI) and the North Pacific T′ index (PI) in this study. The AI–PI correlation also delineates an inverse relationship of upper-tropospheric T′ between Asia and the North Pacific. The AI–PI correlation coefficient in the observation is -0.80, significant at the 99.9% level. Thus, both the EOF and correlation analyses indicate that high (low) T′ over Asia accompanies low (high) T′ over the North Pacific in the upper troposphere. This out-of-phase relationship is also detected in the twentieth century reanalysis V2 products during 1901–2000 and is shown in Figure 2(b), in which the EOF1 mode accounts for 27% of the total variance. The variability of such an out-of-phase relationship in spring conforms to the previous results (Zhao et al., 2010; Zhao et al., 2011) and is named APO (Zhao et al., 2007b).
In the PD simulation (Figure 2(b)), the EOF1 mode accounts for 20% of the total variance and is similar to that from the two reanalysis datasets. The opposite variation in the spring upper-tropospheric T′ between Asia and the North Pacific can be well modelled. Besides, the AI index is negatively correlated to the PI index in the PD simulation, with a correlation coefficient of − 0.77 (significant at the 99.9% level). This is also consistent with the present reanalysis results.
4. Changes of spring APO and associated atmospheric circulation in the mid-Holocene
Figure 1(c) presents the climatological distribution of spring upper-tropospheric T′ in the MH simulation. Clearly, the pattern of T′ in the MH simulation is generally consistent with that in the PD simulation and the reanalysis. The Eurasian continent was dominated by positive values and the North Pacific was predominated by negative values. The EOF1 mode of spring T′ in the MH simulation is shown in Figure 2(c), explaining 20% of the total variance. The simulated EOF1 pattern in the mid-Holocene also resembles that in the PD simulation (Figure 2(b)). These results demonstrate that the APO phenomenon in upper-tropospheric temperature also occurred in spring of the MH.
However, there are some differences in spring APO between the MH and the PD. Figure 3(a) depicts the composite difference of the spring upper-tropospheric T′ between the MH and PD simulations. The negative values over Asia and the positive values over the North Pacific are very notable. These significant differences over the Asian-Pacific region appeared almost throughout the entire troposphere, with the largest difference between 300 and 200 hPa (Figure 3(b)). Therefore, the modelled spring upper-tropospheric T′ was colder over Asia and warmer over the North Pacific in the MH compared with the present, which indicates a weaker zonal thermal contrast between Asia and the North Pacific at that time.
Following Zhao et al. (2007b), we defined an APO index (APOI) as the arithmetic difference between the AI and PI indices, i.e.
The APOI has a significant positive correlation of 0.94 (0.96) with the AI and a significant negative correlation of − 0.95 (−0.96) with the PI for the MH (PD) simulation. Thus, the APOI can well represent variations of the Asian and North Pacific T′ and their difference. Moreover, the APOI is also significantly correlated to the time series of the EOF1 mode, with the correlation coefficient of 0.79 (0.96) in the MH (PD) simulation. In the MH (PD) simulation, the mean values of AI, PI and APOI are 0.98 °C (1.32 °C), − 0.79 °C (−1.03 °C) and 1.77 °C (2.35 °C), respectively. Clearly, these indices in the MH were smaller than those at present, further supporting the result from Figure 3 which shows colder T′ over Asia and warmer T′ over the North Pacific as well as weaker APO in the MH with respect to the present.
Figures 4 and 5 respectively manifest the regressions of the eddy geopotential height H′ (i.e. the departure of the geopotential height from its zonal mean) at 850 and 100 hPa against the spring APOI in the PD and MH simulations. It is found that the regression pattern in two simulations bears general resemblance to each other. In association with a strong phase of spring APO, the lower-tropospheric H′ tends to decrease over Asia and increase over the North Pacific. In the upper troposphere south of 40°N, H′ is inclined to increase over Asia and to decrease over the North Pacific. Therefore, the weakened APO intensity in the mid-Holocene may affect the H′ change over the Asian-Pacific region. As shown in Figure 6, the positive (negative) H′ difference covers Asia and the negative (positive) H′ difference occupies the North Pacific in the lower (upper) troposphere.
Figure 7(a) shows the longitude-height cross section of the spring H′ along 5–40°N in the PD simulation. It is explicit that an upper-tropospheric high (low) and a lower-tropospheric low (high) dominate Asia (the central-eastern Pacific). The composite difference of the vertical structure of spring H′ between two simulations is plotted in Figure 7(b). Over Asia (the central-eastern Pacific), corresponding to the negative (positive) T′ difference shown in Figure 3(b), there are negative (positive) H′ differences in the upper troposphere and positive (negative) H′ differences in the lower troposphere. This result implies that both the upper-tropospheric high (low) and the lower-tropospheric low (high) over Asia (the central-eastern Pacific) were weakened in the mid-Holocene relative to the present.
The variation of horizontal winds between the MH and PD simulations conforms to the H′ change as shown in Figure 6. In the upper troposphere (Figure 8(b)), an anomalous cyclonic circulation over Asia and an anomalous anticyclonic circulation to the south of the North Pacific can be explicitly observed. In the lower troposphere (Figure 8(a)), Asia and the extratropical North Pacific are dominated respectively by an anomalous anticyclonic circulation and an anomalous cyclonic circulation. Meanwhile, anomalous northerly winds prevail over East Asia, unfavourable for the occurrence of local precipitation (Zhou and Zhao, 2010a).
The APO change has been documented to be closely connected with variations of the tropospheric zonal vertical cell over the Asian-Pacific region (Zhao et al., 2010). Figure 9(a) depicts the longitude-height cross section of spring zonal vertical circulation along 5–40°N in the PD simulation. A large-scale zonal vertical cell is very apparent in the troposphere, with the ascending over East Asia and the descending over the North Pacific, similar to that of the reanalysis during summer (Zhao et al., 2010). The composite difference of the zonal vertical circulation between the MH and PD simulations is presented in Figure 9(b). Evidently, anomalous downward motion appears over East Asia, indicating a weaker upward motion of the tropospheric zonal cell in the MH. Meanwhile, anomalous upward motion mainly emerges over the eastern Pacific, indicating a weaker downward motion in the MH.
The changes of the vertical motion associated with spring APO may further induce anomalies of local precipitation. In spring, a southwest–northeast-oriented rainbelt exceeding 2 mm/d appears from southern China to the North Pacific (Figure 10(a)) in the PD simulation, which is in agreement with the observation (Figure not shown). This consistency suggests the capability of the CCSM3 in simulating the spring rainbelt in the East Asia–North Pacific region. The simulated rainfall pattern in the MH (Figure 10(b)) is generally approximate to that in the PD simulation, but with different intensity. As shown in Figure 10(c), corresponding to the weak upward motion in the MH over East Asia, local precipitation generally reduced at that time as compared with the present. In contrast, because of the weakness of spring APO in the MH, anomalous ascending was introduced over most part of the North Pacific, consequently leading to an increase of local precipitation in the MH.
It is not easy to find out spring proxies of atmospheric circulation to support the modelling results from the CCSM3. It is also noted that records associated with dry/wet signal over the Asian-North Pacific region for the springtime of the MH are relatively few, although there are many proxy data for annual mean and summer precipitation. Thus, we compare the CCSM3 results with those from other models of PMIP2 for validating the reliability of the CCSM3 results. Although there are other MH simulations in the PMIP2 archive for a range of models, they did not provide the entire tropospheric temperature data. So, this comparison is emphasized on 200 hPa T′, 850 hPa H′, and precipitation. As seen in Figure 11, the CCSM3 model simulates negative difference over Asia and positive difference over the North Pacific between the MH and PD at 200 hPa. This pattern over the Asian-Pacific region is similar to that of the upper-tropospheric T′ as shown in Figure 3(a) and can also be simulated by most of PMIP2 models, although the quantitative amplitudes and patterns are somewhat different among the models. Similarly, most of the models reproduce the increase (decrease) of 850 hPa H′ (Figure 12) over Asia (the North Pacific) and the decrease (increase) of precipitation in East Asia (a wide area of the North Pacific) (Figure 13) in the MH relative to the present, despite some divergence among different models. These results also agree with that of the CCSM3 simulation. Thus, the consistency of the results between CCSM3 and many other PMIP2 models supports the reliability of the findings in this study.
5. Conclusion and discussion
In this study, we address the change of spring upper-tropospheric teleconnection in the mid-Holocene by using the CCSM3 simulation data. The simulation result shows an out-of-phase relationship in the variability of upper-tropospheric T′ between Asia and the North Pacific in the springtime of the MH, which is consistent with the APO phenomenon of the present climate. Nevertheless, compared with the present climate, the upper-tropospheric T′ in the mid-Holocene was simulated to be colder over Asia and warmer over the North Pacific, indicating a weak APO in the mid-Holocene.
Changes of the atmospheric circulations and climate associated with the spring APO in the mid-Holocene are also investigated in this study. Concurrent with a weaker APO in the springtime of the MH, the eddy geopotential height decreased in the upper troposphere and increased in the lower troposphere over Asia, while it increased in the upper troposphere and decreased in the lower troposphere over the North Pacific. As a result, anomalous descending motion appeared over East Asia, leading to a decrease of local rainfall in the MH. Conversely, anomalous ascending motion emerged over the North Pacific, resulting in more local precipitation in the MH. Therefore, the spring APO teleconnection may well measure the change of atmospheric circulations and precipitation over the Asian-Pacific region in the mid-Holocene.
We also validate the finding in this study by comparing the CCSM3 results with those from other PMIP2 models. The general pattern with the negative (positive) 200 hPa T′ difference, positive (negative) 850 hPa H′ difference and negative (positive) precipitation difference over Asia (a wide area of the North Pacific) between the MH and PD in the CCSM3 simulation can also be simulated by most of PMIP2 models, although the quantitative amplitudes and patterns are somewhat different among these models. This consistency suggests that the results from the CCSM3 simulation are reliable. However, the simulated change of MH precipitation over the Asian-North Pacific region in spring needs to be validated by records retrieved from geological evidence in the future.
The out-of-phase relationship in the variability of upper-tropospheric T′ over the Asian-Pacific region also existed in the summertime of the mid-Holocene (Zhou and Zhao, 2010b). However, the MH APO change was different between spring and summer, showing a pronounced seasonal variation. In contrast to the weakening of the MH APO in spring, the summer APO was intensified in the MH compared with the present, indicative of a stronger summer zonal thermal contrast between Asia and the North Pacific and a stronger East Asian summer monsoon. Such a difference is likely related to the atmospheric response to a seasonal change of solar radiation in the MH. Otto-Bliesner et al. (2006) revealed that negative solar radiation anomalies in spring and positive solar radiation anomalies in summer occurred in the Northern Hemisphere for the MH relative to the present (see their Figure 1). The difference of solar radiation anomalies in spring and summer may force different temperature response, and thus result in different change of the APO intensity in the springtime and summertime of the MH.
This research was supported by the National Basic Research Program of China under Grants 2009CB421407 and 2007CB815901, and the Special Fund for Public Welfare Industry (meteorology) under Grant GYHY200906018. We acknowledge the international modelling groups for providing their data for the analysis, the Laboratoire des Sciences du Climat et de l'Environnement (LSCE) for collecting and archiving the model data. The PMIP2/MOTIF Data Archive is supported by CEA, CNRS, the EU project MOTIF (EVK2-CT-2002-00153) and the Programme National d'Etude de la Dynamique du Climat (PNEDC). More information is available on http://pmip2.lsce.ipsl.fr/ and http://motif.lsce.ipsl.fr/.