What drives the global summer monsoon over the past millennium?
- 1k Downloads
The global summer monsoon precipitation (GSMP) provides a fundamental measure for changes in the annual cycle of the climate system and hydroclimate. We investigate mechanisms governing decadal-centennial variations of the GSMP over the past millennium with a coupled climate model’s (ECHO-G) simulation forced by solar-volcanic (SV) radiative forcing and greenhouse gases (GHG) forcing. We show that the leading mode of GSMP is a forced response to external forcing on centennial time scale with a globally uniform change of precipitation across all monsoon regions, whereas the second mode represents internal variability on multi-decadal time scale with regional characteristics. The total amount of GSMP varies in phase with the global mean temperature, indicating that global warming is accompanied by amplification of the annual cycle of the climate system. The northern hemisphere summer monsoon precipitation (NHSMP) responds to GHG forcing more sensitively, while the southern hemisphere summer monsoon precipitation (SHSMP) responds to the SV radiative forcing more sensitively. The NHSMP is enhanced by increased NH land–ocean thermal contrast and NH-minus-SH thermal contrast. On the other hand, the SHSMP is strengthened by enhanced SH subtropical highs and the east–west mass contrast between Southeast Pacific and tropical Indian Ocean. The strength of the GSMP is determined by the factors controlling both the NHSMP and SHSMP. Intensification of GSMP is associated with (a) increased global land–ocean thermal contrast, (b) reinforced east–west mass contrast between Southeast Pacific and tropical Indian Ocean, and (c) enhanced circumglobal SH subtropical highs. The physical mechanisms revealed here will add understanding of future change of the global monsoon.
KeywordsGlobal summer monsoon precipitation Northern hemisphere summer monsoon precipitation Southern hemisphere summer monsoon precipitation Past millennium Centennial time scale Multi-decadal time scale
Monsoon is a forced response of the coupled atmosphere–ocean–land system to annual variation of insolation. It is characterized by an annual reversal of prevailing surface winds and a contrast between wet summer and dry winter. The present analysis focuses on monsoon precipitation as it is a more important variable than the winds to affect the environment and society.
Based on monsoon rainfall characteristics, monsoon domains can be delineated by the regions where the annual range (local summer mean minus winter mean) of precipitation rate exceeds a threshold of 2.0 mm/day and the local summer precipitation exceeds 55 % of the annual total (Wang and Ding 2006; Liu et al. 2009). Here summer is defined as May through September (MJJAS) for the NH and November through next March (NDJFM) for the SH. The defined monsoon precipitation domain is shown in Fig. 2.
The climate variability of the global monsoon can be measured by the global summer monsoon precipitation (GSMP), which is defined as the local summer precipitation in the global monsoon domain, i.e., MJJAS precipitation in the NH and the ensuing NDJFM precipitation in the SH. Since the annual range in monsoon regions is dominated by the local summer precipitation, the GSMP represents approximately the amplitude of the annual variation of precipitation.
Future climate change involves both internal and forced variability in the climate system. To distinguish contributions to climate change from internal variability and forced variability, it requires investigation of climate variations on a longer period, such as in the last millennium. Unfortunately, such an analysis is impeded by substantial limitations of the availability and quality of global historical records.
An alternative and effective approach is to probe causes of the climate variability simulated by coupled climate models that can reproduce reliable long-term climate variations. The underlying assumption is that the models’ physics is robust enough for providing adequate physical understanding of the causes of the past climate variations. This type of millennial simulations has been constructed using complex coupled models, including the HadCM2 model (Johns et al. 1997), CSIRO model (Vimont et al. 2002), ECHO-G model (Rodgers et al. 2004), the NCAR-CSM (Mann et al. 2005; Ammann et al. 2007) and others. Various aspects of the climate variability simulated by the ECHO-G model have been examined, including the intrinsic internal variability, temperature variation, ENSO, AMOC (Atlantic meridional overturning circulation), NAO (North Atlantic Oscillation) and monsoons (Zorita et al. 2003, 2005; Gonzalez-Rouco et al. 2003; Gouirand et al. 2007a, b; Rodgers et al. 2004; Min et al. 2005a, b; Wagner et al. 2005; Liu et al. 2009). The results gained from these studies have generally built on the model’s credibility for comprehension of pertinent physical processes responsible for the simulated climate variability.
Thus far, no analysis of the secular changes of GSMP has been carried out in terms of any climate model’s millennial integration except a recent work by Liu et al. (2009). They found that the strength of the global-average summer monsoon precipitation in the forced run exhibits a minimum during the Little Ice Age (LIA) and a maximum during Medieval Warm Period (MWP), which follows the natural variations in the total amount of effective solar radiative forcing. The notable strengthening of the global summer monsoon rainfall in the last 30 years of the forced simulation (1961–1990) is unprecedented and owed in part to the increase of atmospheric CO2 concentration. However, the spatial patterns, the hemisphere differences in local summer monsoons and the associated mechanisms were not studied.
Here we focus on addressing the following questions: What are the spatial–temporal structures of the global scale local summer monsoon precipitation (GSMP) change on the timescales longer than a decade over the past millennium? Can we distinguish the forced response and internal variability of the GSMP? What are the driving mechanisms for the GSMP as well as the NHSMP and SHSMP? Are there any difference in the driving mechanisms between the NHSMP and SHSMP?
2 The model and validation
2.1 Model and experiments
The numerical model used for millennial integration is the ECHO-G coupled climate model (Legutke and Voss 1999), which consists of the spectral atmospheric model ECHAM4 (Roeckner et al. 1996) and the global ocean circulation model HOPE-G (Wolff et al. 1997). The description of the model physics and performance was given in details by Zorita et al. (2003). Here for convenience of the readers, we provide a brief summary. The model configuration has 19 vertical levels in the atmosphere and 20 levels in the ocean, and horizontal resolutions are approximately 3.75° (atmosphere) and 2.8° (ocean) in both latitudes and longitudes. The ocean model HOPE-G has a grid refinement in the tropical regions, where the meridional grid point separation reaches 0.5°. To enable the coupled model to sustain a simulated climate near to the real present day climate with minimal drift, both heat and fresh-water fluxes between atmosphere and ocean are modified by adding a constant (in time) field of adjustment with net-zero spatial average (Roeckner et al. 1996; Wolff et al. 1997).
The initial conditions of the ERIK simulation were taken from year 100 of the control run. Those initial conditions are, however, representative of present-day rather than pre-industrial climate and the experimental design therefore included a 30-year adjustment period during which the control run forcing was linearly reduced until it matched the forcing imposed around 1000 AD, followed by a 50-year period with fixed forcing to allow the model’s climate to readjust to the modified forcing. The ERIK simulation then proceeded from the conditions at 1000–1990 AD. Note that the uncertainty in the initial conditions, in turn, could potentially influence the relationship between applied forcing and simulated response in the first 100–200 years. This issue has been addressed by Zorita et al. (2007).
2.2 Validation of the model precipitation climatology
How the model responds to annual variation of the insolation is critical for model evaluation because we examine how the model responds to long-term variability in insolation forcing. The performance of ECHO-G in modeling the patterns and errors of the climatological annual mean and annual cycle of precipitation has been assessed by Liu et al. (2009). Their results indicate that the simulated precipitation climatology in ERIK run is comparable to those assimilated climatology in NCEP-2 reanalysis. The overall agreement adds confidence to our further analysis of the decadal-centennial precipitation variability using the outputs generated by the ERIK run. Despite these successes, the simulated annual mean precipitation has notable biases in East Asia monsoon region, subtropical South Pacific and South Atlantic convergence zones, and the Mexican-North American monsoon regions (Liu et al. 2009). These biases are also common in most of coupled models in simulating annual mean precipitation (Lee et al. 2010).
2.3 Statistical significance test for the difference of the regression coefficients
3 The forced and internal modes of the decadal GSMP
Can we distinguish forced and internal variability of the GSMP? The separation is possible if the forced response and internal variability have distinct spatial patterns and temporal behaviors. For this reason, we examine the principal modes of the GSMP variability by empirical orthogonal function (EOF) analysis of the decadal mean precipitation in monsoon domains over the past millennium. The first and second modes accounts for 9.5 and 5.4 % of the total variance, respectively, and they are significantly separated from each other and from the rest higher modes using the method of North et al. (1982).
4 Differences between the NHSMP and SHSMP variations
Figure 6b shows that the change of GSMP over the past millennium follows that of the GMT closely (r = 0.87). This in-phase variation occurs primarily on the centennial-millennial time scale because the interdecadal variations are dominated by the internal feedback process which does not contribute significantly to global mean quantities. The GMT decreases from 1140 to 1700 AD and then increases from 1700 to 1990 AD. Similarly, the GSMP decreases also roughly from 1140 to 1700 AD and rises from 1700 to 1990 AD. The rapid increase of GMT since around 1850 seems to be associated with the rapid increase of CO2 concentration.
Note that the global monsoon is the dominant mode of the annual variation of the tropical precipitation and circulation (Wang and Ding 2008) and the GSMP intensity measures change of the annual cycle of the coupled climate system. Therefore, the in-phase relationship between the GMT and GSMP indicates that global warming is accompanied by amplification of the annual cycle of the Earth’s climate system. Furthermore, the GMT measures the annual mean climate but the GSMP measures the annual cycle of the climate system. Thus, the GSMP provides a sensible and complementary measure of global climate change in terms of a key hydrological variable.
Why is the NHSMP more sensitive to GHG than to SV forcing? It is conceivable that the GHG forcing creates larger land–ocean thermal contrast than SV forcing. This has been tested by calculation of the thermal contrast changes in preindustrial period (MWP minus LIA) and the industrial period (Present minus LIA). The results show that the NH land–ocean thermal contrast is 0.18 °C for MWP minus LIA and 0.48 °C for Present minus LIA, indicating that GHG forcing is indeed more effective than SV forcing in creating land–ocean thermal contrast, and thus favors increased NHSMP.
5 Factors controlling the NHSMP, SHSMP, and GSMP
Figure 9 indicates that a stronger NHSMP is related to (a) an increased land–ocean thermal contrast (warm land and cold ocean in the NH), and (b) an increased hemispheric thermal contrast (warm NH and cold SH). Increased land–ocean thermal contrast can enhance monsoon lows and the associated moisture convergence (Liu et al. 2009); and warmer NH generates cross-equatorial pressure gradients that drive low-level cross-equatorial flows from SH to NH, again strengthening the NH monsoon (Loschnigg and Webster 2000). The combined effects of the land–ocean and NH-SH thermal contrasts yield a strong control of the NHSMP variability on decadal-centennial time scales.
Different from the NHSMP, the SHSMP is not significantly related to the land–ocean and hemispheric thermal contrasts (figure not shown). Instead, it is associated with (a) an east–west thermal contrast-induced SLP difference between the southeastern Pacific and tropical Indian Ocean, and (b) the circumglobal SH subtropical high strength measured by the SLP averaged between 20°S and 40°S (the ridge lines of the SH subtropical high locate at about 30°S) (Fig. 10). The SHSMP is enhanced when pressure rises in the southeastern Pacific and drops in the tropical Indian Ocean. The increased east–west SLP gradient induces a westward air flow and moisture transport, which enhances moisture convergence over the southern African-southwest Indian Ocean and the Australian monsoon region. The enhanced SH subtropical high increases the trade winds in SH oceans, also strengthening the convergence of air mass and moisture in the SH monsoon low pressure regions.
The GSMP is the sum of the NHSMP and SHSMP, so it is affected by the common factors that control the NH and SH summer monsoons. Analysis shows that the GSMP is most closely associated with (a) the east–west thermal contrast-induced SLP difference between the southeastern Pacific and tropical Indian Ocean (r = 0.88), (b) the land–ocean thermal contrast over the globe between 40°S and 60°N (r = 0.84), and (c) the circumglobal SH subtropical high strength between 40°S and 20°S (r = 0.88) (Fig. 11). The enhanced “southeastern Pacific cooling-Indian Ocean warming” induces rising SLP in the eastern Pacific, which enhances the trade winds, transporting and converging moisture into the eastern hemisphere monsoon regions, which not only favor enhancement of the SH monsoon (Fig. 10) but also NH monsoon through enhancing cross equatorial flows. The increased land–ocean thermal contrast enhances monsoon low and associated moisture convergence, and primarily favors the NH summer monsoon precipitation. The enhanced SH subtropical high strengthens not only the SHSMP (Fig. 10) but also the NHSMP by increasing the pressure gradient between the NH and SH. The combined effects of the east–west and the land–ocean thermal contrasts, as well as the SH subtropical high strength result in a strong control of the GSMP.
6 Concluding remarks
The GSMP measures the annual range of precipitation averaged in the GM domain, thereby providing a useful parameter to quantify the change in the annual cycle of the earth climate. This new measure also provides important information about the change in the global monsoon precipitation, which is a climate variable that is far more relevant for food production and water supply than the mean temperature change. Note that the GMT measures change of annual mean climate in terms of temperature, whereas GSMP measures change of the annual cycle in terms of precipitation. They are complementary in nature. The linkage between the GMT and GSMP indicates the close relationship between the global warming and amplification of the annual cycle of the climate system.
We have shown that forced response of the GSMP has a distinct spatial–temporal structure from that due to internal feedback processes. Two prominent patterns of GSMP variability are identified. The leading pattern has a dominant centennial-millennial variation and features a nearly uniform increase of monsoon precipitation across all regional monsoons. This pattern is a forced response to the changes in effective solar-volcanic (SV) radiation and GHG concentration. The second pattern is associated with a multi-decadal oscillation in the central Pacific sea surface temperature (SST), representing an internal feedback mode.
An important finding is that the NHSMP and SHSMP have notable differences in terms of their driving mechanisms. First, the NHSMP responds to GHG forcing more sensitively than the SV forcing, while the SHSMP responds to the natural solar-volcanic radiative forcing more sensitively than the NHSMP does. Second, the enhanced NHSMP is primarily driven by warm land-cold ocean in NH and warm NH-cold SH, while the enhanced SHSMP is primarily caused by enhanced east–west thermal contrast over the SH Indo-Pacific warm pool and the SH subtropical high strength. The GSMP is driven by the factors that commonly control both the NH and SH summer monsoons, including the east–west thermal contrast between the Southeast Pacific and tropical Indian Ocean, the global land–ocean thermal contrast, and the circumglobal SH subtropical high strength.
These results obtained from the present pilot study carried out by using ECHO-G millennial runs should be compared with available proxy data and tested by using other models and by analyzing model outputs on other time scales. These include examination of PMIP-3 ensemble model outputs for past millennium, last glacial maximum and mid-Holocene periods.
The physical mechanisms leant from this study will also add understanding of future change of the global monsoon. For instance, the climate models’ future projections under increasing GHG forcing display an increased land–ocean thermal contrast and the contrast between warmer NH and colder SH. These two factors in principle may favor enhancement of the NHSMP, and the GSMP as well.
We would like to thank Dr. Eduardo Zorita for providing the modeling results and helpful comments. This work was supported by the National Basic Research Program and Natural Science Foundation of China (Award #. 2010CB950102, XDA05080800, 2011CB403301, 2010CB833404, and 40890054) (JL and BW), the US NSF award #AGS-1005599 (BW), and the Global Research Laboratory (GRL) Program from the National Research Foundation of Korea grant # 2011-0021927 (KJH, BW, JYL, JGJ). BW, SYY and JYL acknowledge support from the International Pacific Research Center which is funded jointly by JAMSTEC, NOAA, and NASA. This is publication no. 8589 of the School of Ocean and Earth Science and Technology and publication no. 868 of the International Pacific Research Center.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
- Gouirand I, Moberg A, Zorita E (2007a) Climate variability in Scandinavia for the past millennium simulated by an atmosphere-ocean general circulation model. Tellus 59A:30–49Google Scholar
- Legutke S, Voss R (1999) The Hamburg atmosphere-ocean coupled circulation model ECHO-G. German climate computer center (DKRZ). Technical Report, vol 18Google Scholar
- Min SK, Legutke S, Hense A, Kwon WT (2005a) Internal variability in a 1000-year control simulation with the coupled climate model ECHO-G—I. Near-surface temperature, precipitation and mean sea level pressure. Tellus 57A:605–621Google Scholar
- Min SK, Legutke S, Hense A, Kwon WT (2005b) Internal variability in a 1000-year control simulation with the coupled climate model ECHO-G—II. El Niño Southern Oscillation and North Atlantic Oscillation. Tellus 57A:622–640Google Scholar
- Roeckner E, Arpe K, Bengtsson L, Christoph M, Claussen M, Dumenil L, Esch M, Giorgetta M, Schlese U, Schulzweida U (1996) The atmospheric general circulation model ECHAM4: model description and simulation of present-day climate. Max Planck Institute for Meteorology, Technical Report 218Google Scholar
- Wolff JO, Maier-Reimer E, Legutke S (1997) The Hamburg Ocean primitive equation model. German climate computer center (DKRZ). Technical Report 13Google Scholar