Spatial variability of Holocene changes in the annual precipitation pattern: a model-data synthesis for the Asian monsoon region
- First Online:
This study provides a detailed analysis of the mid-Holocene to present-day precipitation change in the Asian monsoon region. We compare for the first time results of high resolution climate model simulations with a standardised set of mid-Holocene moisture reconstructions. Changes in the simulated summer monsoon characteristics (onset, withdrawal, length and associated rainfall) and the mechanisms causing the Holocene precipitation changes are investigated. According to the model, most parts of the Indian subcontinent received more precipitation (up to 5 mm/day) at mid-Holocene than at present-day. This is related to a stronger Indian summer monsoon accompanied by an intensified vertically integrated moisture flux convergence. The East Asian monsoon region exhibits local inhomogeneities in the simulated annual precipitation signal. The sign of this signal depends on the balance of decreased pre-monsoon and increased monsoon precipitation at mid-Holocene compared to present-day. Hence, rainfall changes in the East Asian monsoon domain are not solely associated with modifications in the summer monsoon circulation but also depend on changes in the mid-latitudinal westerly wind system that dominates the circulation during the pre-monsoon season. The proxy-based climate reconstructions confirm the regional dissimilarities in the annual precipitation signal and agree well with the model results. Our results highlight the importance of including the pre-monsoon season in climate studies of the Asian monsoon system and point out the complex response of this system to the Holocene insolation forcing. The comparison with a coarse climate model simulation reveals that this complex response can only be resolved in high resolution simulations.
KeywordsAsian monsoon Holocene Precipitation Climate modelling Moisture reconstructions
Monsoon systems primarily form due to seasonal and latitudinal differences in the incoming solar radiation. The different heat capacities of the continent and ocean lead to strong land-sea thermal contrasts (Webster et al. 1998). As consequence, large-scale pressure gradients are generated inducing strong low-level atmospheric wind circulations that reverse seasonally according to the change in insolation. The strength of the monsoonal circulation is susceptible to variations in all components of the climate system i.e. of those on land, in the atmosphere and in the oceans. Monsoon systems are therefore characterised by a strong temporal variability covering multi-millennial to intraseasonal timescales (e.g. Lau et al. 2000; Wang 2006; Ding 2007).
The Asian monsoon system is the strongest monsoon system of the world (Clift and Plumb 2008). It incorporates the East Asian and the Indian monsoon, which are independent sub-systems showing different characteristics and variability, but also strongly interact (Wang et al. 2001; Ding and Chan 2005). These differences are at least partly related to the different land–ocean configurations (including topography) in the sub-monsoon domains (Wang et al. 2003).
The Indian summer monsoon is characterised by the meridional pressure-gradient between the monsoon trough in North India and the Mascarene high causing an interhemispheric circulation (Wang et al. 2003; Wang 2006). The strength of this circulation and the onset and maintenance of the Indian summer monsoon are to a large part determined by climate processes on the Tibetan Plateau that forms an elevated heat source for the atmosphere in late spring and summer and enhances the thermal contrast between the land and ocean in the Indian monsoon sector (e.g. Ye and Wu 1998; Wu et al. 2007). The seasonal March of the Indian monsoon rainband is strongly connected with the movement of the Intertropical Convergence Zone (ITCZ) and reaches its northernmost position in mid-July (Wang et al. 2003). More than 80 % of the annual precipitation in the Indian subcontinent originates from the summer monsoon (Wang et al. 2001). Rainfall peaks early in summer (June–July, Wang et al. 2003).
In contrast, the East Asian summer monsoon is controlled by the zonal pressure-gradient between the monsoon trough on the Asian continent and the western Pacific subtropical high (Webster et al. 1998). As a result of the east–west thermal contrast, a planetary subtropical front is formed (Meiyu-Baiu frontal system) whose seasonal March determines the propagation of the East Asian summer monsoon rainbelt (e.g. Ding 2007). The subtropical front reaches the northernmost position during early July and is separated from the ITCZ, which forms a second rainband in East Asia (tropical rainbelt) affecting mainly the western North Pacific monsoon. Rainfall associated with the ITCZ over the East Asian sector peaks during August and only affects the area south of 25°N (Wang et al. 2003). The interplay of the mid-latitude westerlies, the subtropical circulation around the western Pacific subtropical high and the tropical circulation including the Indian monsoon flow makes the East Asian monsoon domain unique and enables the occurrence of many significant weather and climate phenomena (Ding 2007). In contrast to other monsoon regions, the precipitation in large parts of the East Asian monsoon domain is not temporally confined to the summer monsoon season. According to Yihui and Zunya (2008), spring precipitation accounts for approximately 32 % of the annual total in the middle and lower reaches of the Yangtze River (110–122°E, 28–32°N) and is also very dominant in the Huaxi region (100–110°E, 30–35°N). In South China (110–122°E, 20–28°N), spring and summer precipitation contribute to the annual total in equal parts (35 and 38 %, respectively). However, precipitation in North (115–125°E, 35–42°N) and North East China (122–135°E, 40–52°N) clearly peaks in summer as expected for monsoon regions.
At mid-Holocene, differences in the Earth’s orbital configuration compared to present-day lead to differences in the incoming solar radiation (Berger 1978). During summer (winter), the insolation was increased (decreased) by approx. 5 % on the Northern Hemisphere yielding an enhancement of the seasonal cycle. Climate modelling studies suggest a general strengthening of the Asian summer monsoon circulation as response to this insolation forcing that can mainly be explained by an intensification of the land-sea thermal gradient (e.g. Braconnot et al. 2007a, 2008; Marzin and Braconnot 2009; Wang et al. 2010a). Precipitation is assumed to be increased in most parts of the continental monsoon domain during the mid-Holocene, but the analysis is often confined to the summer season (June–August, e.g. Wang et al. 2010a) which may neither portray the entire summer monsoon season nor the annual mean precipitation in this region.
Feedbacks in the climate system such as interactions of the atmosphere and the ocean or the vegetation have been shown to modify the initial response to the insolation forcing. The strength of these interactions reveals regional dissimilarities in the Asian monsoon domain as well as contradictory contributions to the Holocene climate change in different climate models (see Dallmeyer et al. 2010 and references therein). To assess the performance of the different models, the model results have to be validated to paleoclimate reconstructions.
Climate reconstructions confirm the increase in precipitation in most regions of the Asian monsoon domain during early and mid-Holocene (e.g. Fleitmann et al. 2003; Wang et al. 2005; Herzschuh et al. 2006; Prasad and Enzel 2006) and suggest differences in the precipitation response to the insolation forcing between the Indian and East Asian monsoon region (An et al. 2000; Maher and Hu 2006; Maher 2008). However, most of the proxies do not resolve seasonal signals but rather record an aggregated annual signal. Therefore, the proxies cannot a priori provide information on seasonal climate phenomena such as the summer monsoons. For example, the precipitation signal in vegetation-proxies is smoothed by the soil moisture storage capacity. Lake water level proxies from closed lakes record the total accumulated precipitation relative to evaporation. However, by analogy to the present summer precipitation dominance these proxy-records have often served as proof for the simulated intensification of the Asian summer monsoon. So far, the reconstructions have lacked a consistent age control and revealed large disparities in the sample resolution that might have influenced the interpretation of these records and their applicability in comparative studies of climate modelling results and reconstructions. Recently, Wang et al. (2010b) have provided a synthetic review of the spatial and temporal pattern of moisture evolution in monsoonal central Asia that is based on more than 90 synchronised proxy records.
In this study, we compare results of high-resolution global climate model simulations with these synchronised reconstructions for the mid-Holocene time-slice. Using the climate model results, we (1) quantify changes in the summer monsoon characteristic between the mid-Holocene and present-day, (2) investigate how seasonal precipitation changes affect the simulated annual precipitation signal, (3) identify atmospheric mechanisms leading to the reconstructed spatial moisture pattern and (4) analyse the processes causing the differences in the precipitation response between the Indian and East Asian monsoon region. With this study, we provide for the first time a comprehensive and consistent comparison of reconstructions and climate model results for the mid-Holocene Asian monsoon climate.
2.1 Model setup and experiments
To analyse differences in climate between the mid-Holocene (6,000 years before present, referred to as 6k) and present-day (0k), we performed time-slice experiments with the model ECHAM5/JSBACH. This model consists of the general circulation model for the atmosphere ECHAM5 (Roeckner et al. 2003) and the land surface scheme JSBACH (Raddatz et al. 2007) that also included a dynamic vegetation module in our simulations (Brovkin et al. 2009). The models ran with the spectral resolution T106L31. This corresponds to a longitudinal distance of approx. 1.125° and 31 levels in the vertical.
In this study, we conducted two experiments, one with mid-Holocene orbital configuration (AV6k), one with present-day orbital configuration (AV0k) prescribed. In AV6k and AV0k, monthly mean sea surface temperature and sea-ice distribution were prescribed according to their mid-Holocene and pre-industrial distributions, respectively. To account for interannual variability, these values were taken from 120 years of a coarse resolution simulation. This simulation has been performed with the comprehensive Earth system model ECHAM5/JSBACH-MPIOM (Fischer and Jungclaus 2011), which included the general circulation model for the ocean MPIOM (Marsland et al. 2003).
During both simulations, atmospheric composition was fixed at pre-industrial values, i.e. CO2-concentration was set to 280 ppm. ECHAM5/JSBACH had been brought to quasi-equilibrium climate state (spin-up of approx. 200 years) before it ran for additional 30 years (analysis period). All results and plots are based on these 30 years of analysis period.
The model results have been tested against climate observations and reanalysis data to prove that the model captures the major structure of the present-day global and regional climate. Present-day annual and seasonal precipitation maps as well as a discussion of the major precipitation differences to observational datasets are provided in the “Appendix”.
2.2 Calculation of the monsoon onset and withdrawal period
Past moisture changes inferred from proxy studies in eastern and southern Asia were hitherto generally assumed to originate almost exclusively from changes in the summer monsoon as this circulation pattern is responsible for most of the present-day annual precipitation in these areas; also paleoclimate modelling studies focus often on the summer season (e.g. Wei and Wang 2004; Li et al. 2009; Wang et al. 2010a; Zhou and Zhao 2010). For the purpose of simplification, the summer monsoon season is usually represented by climate means of the months June to August or September and not explicitly calculated in paleoclimate studies. We are not aware of previous modelling studies assessing changes in the summer monsoon onset, withdrawal, length and related precipitation between the mid-Holocene and present-day in detail for the entire Asian monsoon region.
The period in which rpi reaches values above 5 mm/day represents the monsoon season. Accordingly, the onset (withdrawal) is defined as the pentad in which rpi exceeds (falls below) 5 mm/day. This method is a good approximation of the observed present-day monsoon propagation.
2.3 Reconstructions-based moisture index
The moisture signal from the previously published records (mostly based on lake sediment analyses) were transferred into a three-scale moisture index categorising the mid-Holocene climate as ‘wetter than’, ‘similar as’, or ‘drier than’ at present-day. Generally, the moisture indices are relative, in a semi-quantitative sense for each individual case and therefore can only be compared qualitatively to the model results. For detailed information concerning the methods and original results for most sites we refer to Wang et al. (2010b). Meanwhile, during producing the moisture index, we are aware of the following potential uncertainties:
Chronology: We attempted to construct comparable age-depth models among the records based on consistent calibration methods. However, the potentially varying dating uncertainties, especially the so-called “carbon reservoir effect” for the lacustrine records still need to be considered with cautions.
Various moisture indicators: We drew the moisture information from various proxies (e.g. pollen, ostracods, stable isotope measurements, element ratios), with different indications for the moisture changes, e.g. reconstructed precipitation, lake level changes, moisture balance changes and even the qualitative descriptions on the dry/wet conditions, leading to the potentially different responses to the real climate changes.
Seasonal variations in moisture: Considering the rough information deduced from the proxy data, it is impossible to acquire the seasonal variations, limiting the comparability of the index and the modelling results (Wang et al. 2010b). Even more problematic, the single proxies may represent moisture signals of different seasons of the year and most proxies integrate over a certain or even varying period of the year e.g. lake level proxies or vegetation proxies.
3.1 Comparison of the simulated annual precipitation pattern with proxy-based moisture reconstructions
3.2 Simulated change in summer monsoon season characteristics
Following the monsoon season definition of Wang and LinHo (2002), we assess if the difference in summer monsoon precipitation is responsible for the difference in the annual total between the mid-Holocene (6k) and the present-day (0k) simulation.
Related to the shifts in the onset and withdrawal period, the summer monsoon season length is different at 6k and at 0k (Fig. 4c). In large parts of the continent, the monsoon season is prolonged by up to 10 pentads and the monsoon related precipitation is enhanced during the mid-Holocene. Exceptions are parts of Middle and North East India, Indochina and Bangladesh which experience a shorter monsoon season by approx. 2–8 pentads and less monsoonal precipitation in 6k. These are approximately also the areas showing much less annual mean precipitation in the mid-Holocene simulation than in the present-day one, but changes in summer monsoon precipitation cannot fully explain the simulated annual mean precipitation signal (Fig. 2), particularly in the East Asian monsoon region (e.g. Central Eastern China).
Please notice that we performed the seasonal analysis based on the modern calendar. Due to the precession, the starting date and length of the seasons change in time (cf. Joussaume and Braconnot 1997). Taking astronomical fixed-points (equinoxes and solstice) as definition of the seasons, the summer season, for instance, is shorter by 4 days and the autumn season starts earlier at mid-Holocene compared to present-day. As we discuss the annual precipitation cycle in its entirety, the shift in calendar should not affect the conclusions drawn in this study.
4.1 Insolation differences and regional modes of seasonal precipitation
Generally, the reconstructions and the model results show a rather chessboard like, patchy and non-systematic moisture change in the Asian monsoon region. This patchiness underlines the complexity of the Asian monsoon system and it exposes the complexity of the response of this system to the Holocene insolation change revealing differences in the individual sub-regions of the Asian monsoon domain. The cause of this different behaviour can probably be found in the different nature of the monsoon systems. The Indian monsoon is determined by meridional temperature and pressure gradients. Therefore the Holocene insolation change, which is zonally uniform but reveals meridional variations, has a different effect on the Indian monsoon than on the East Asian monsoon that is formed by zonal temperature and pressure gradients. In the Indian monsoon region, not only the incoming solar radiation but also the meridional insolation gradient (and therewith the energy gradient) is enhanced during summer and reduced during the other seasons (cf. Fig. 1). In the East Asian monsoon domain, the incoming solar radiation is enhanced during mid-Holocene summer and reduced in the other seasons, but the modification of the zonal pressure gradient is only related to the different response of the ocean and the continent to the insolation forcing due to their different heat capacities. Therefore, the Indian and East Asian monsoon systems are suggested to respond differently to the Holocene insolation forcing with respect to their dynamical change in the different seasons.
decreased mid-Holocene precipitation in spring and increased mid-Holocene in summer (e.g. NE China, CE China, SW China; these are all areas located in the East Asian monsoon region)
increased mid-Holocene precipitation during summer (e.g. N India, CTP; these are areas located in the Indian monsoon region)
decreased mid-Holocene precipitation in spring and summer (e.g. Indochina; this is an area in the transition zone of both monsoons)
The annual mean precipitation change in all considered areas is, thus, mostly determined by the sum of precipitation changes in spring and summer. In the following, mechanisms leading to the precipitation change in these two seasons are investigated in more detail.
4.2 Mechanisms causing differences in the simulated spring precipitation
4.3 Mechanisms causing differences in the simulated summer precipitation
4.4 Implications for the moisture reconstructions
4.5 Comparison with other climate modelling simulations
Overview on the PMIP2 simulations including an atmosphere, ocean and dynamic vegetation model
ECBILT CLIO VECODE
2.5° × 3.75°, L19
Compared to the high resolution simulation, T31AOV shows a much clearer annual precipitation change in the Asian monsoon domain with drier mid-Holocene conditions in central-eastern and South China and a wetter climate elsewhere (Fig. 13a). This uniform precipitation signal cannot represent the complex signal seen in reconstructions and points out the added value of the high resolution in AV6k and AV0k. In the coarse resolution, the model is probably able to capture the large-scale response of the climate to the insolation forcing, but cannot represent the regional dynamics.
the wetter mid-Holocene climate in Northern India (3 out of 5 models). The other two models [FOAM-LPJ (Fig. 13e) and ECBILT-CLIO-VECODE (Fig. 13d)] also show regions with drier or similar climate in mid-Holocene compared to present-day.
the increased precipitation on the central southern Tibetan Plateau (around 90°E, 30°N), this is shown by all 5 models.
the existence of regions in eastern China receiving less precipitation during mid-Holocene; only UBRIS-HADCM3M2 (Fig. 13f) suggests increased precipitation in the entire eastern China.
the wetter climate in Northeast China (ca. 110–120°E, 38–42°N, 3 out of 5 models)
In northern India, nearly all models simulate the seasonal signal referred to as mode II in our study, showing a strong increased precipitation during mid-Holocene summer and no response in the other seasons (Fig. 14a). FOAM-LPJ and to a minor extent ECBILT-CLIO-VECODE simulated decreased spring precipitation also for northern India during the mid-Holocene compared to present-day.
5 Summary and conclusion
A high resolution numerical experiment conducted with the general circulation model ECHAM5/JSBACH by applying mid-Holocene insolation forcing has been compared to synchronised moisture reconstructions for the Asian monsoon region. Overall the model results and the reconstructions are in good agreement. They both show wetter conditions in large parts of the Indian monsoon region and a partly drier and partly wetter climate in East Asia. Especially in central-eastern China, reconstructions suggest a very inhomogeneous mid-Holocene to present-day annual total precipitation signal which indicates the complexity of the Asian monsoon dynamics. The model is able to capture this patchy signal and attributes the inhomogeneity to the different responses of the East Asian and Indian monsoon system to the insolation forcing. The East Asian monsoon is determined by a zonal pressure gradient, the Indian monsoon depends on a meridional pressure gradient. Therefore, the zonally uniform insolation change has different effects on these two monsoon sub-systems. Large parts of the Indian monsoon region receive most of the annual total precipitation during the summer monsoon season. In these regions, the difference in simulated annual rainfall between mid-Holocene and present-day can be attributed to changes in the summer monsoon season length and in the monsoon related precipitation, i.e. a prolongation and intensification of the summer monsoon in North India and a shortening and weakening of the summer monsoon in Bangladesh and Indochina. In large parts of Eastern China, present-day precipitation is not confined to the summer monsoon season, but also falls during spring. Therefore, the Holocene moisture change also depends on the response of the spring precipitation to the seasonal insolation forcing and, hence, on changes in the mid-latitude westerly wind circulation. The sign of the simulated annual total precipitation difference between mid-Holocene and present-day is determined by the balance of decreasing spring and increasing summer monsoon precipitation as response to the mid-Holocene insolation forcing. This may also explain the inhomogeneity in the reconstructions, since the ratio of spring to summer precipitation change could be less affected by the large-scale circulation but rather be sensitive to the local environment such as the orography that is very complex in the Asian monsoon region.
The coupled atmosphere–ocean-vegetation model simulations performed within the Paleoclimate Modelling Intercomparison Project Phase II (PMIP2) show large discrepancies among each other and have difficulties to capture the complexity of the reconstructed moisture signal. They cannot represent the inhomogeneity of the signal which might be a problem of the numerical resolution used in the PMIP2 simulations. Using coarse spatial resolutions, the models can probably simulate the large-scale response of the monsoon system to the insolation forcing, but not resolve the regional response. The conclusion that the sign of the annual total precipitation change depends on the balance of the spring and summer signals is confirmed by the PMIP2 simulations.
This study reveals the added value of using high numerical resolutions in climate modelling studies of the Asian monsoon system. It furthermore shows the importance of extending the analysis period of monsoon climate change to other seasons and points out the fact that paleoclimate moisture reconstructions cannot simply be interpreted as indicator for the summer monsoon strength and monsoon related precipitation. The variations in the Earth’s orbit lead to changes in the seasonal insolation distribution and hence affect the rainfall in all seasons. The relevance of the non-monsoon seasons for the annual total precipitation might change with time. Therefore, paleoclimate moisture reconstructions in the Asian monsoon region should not simply be correlated and explained with variations in summer insolation.
We acknowledge the international modeling groups participating in PMIP2 for providing their data for analysis and the Laboratoire des Sciences du Climat et de l’Environnement (LSCE) for collecting and archiving the model data. The PMIP2 Data Archive is supported by CEA, CNRS and the Programme National d’Etude de la Dynamique du Climat (PNEDC). The analyses were performed using the final version of the database. More information is available on http://pmip2.lsce.ipsl.fr/. We would like to thank T. Crueger, MPI-M, for constructive discussion and two anonymous reviewers for the helpful comments that improved our manuscript. This research is part of the priority program INTERDYNAMIK funded by the German Research Foundation (DFG). The high resolution climate model simulations have been performed at the German Climate Computing Center (DKRZ) in Hamburg, Germany.
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.
Appendix: Simulated present-day precipitation pattern
The comparison of the Indian transects exhibits a strong rainfall deficit in the monsoon region nearly all over the year (Fig. 17a, c, e). Particularly in the summer season, the mean pentad rainfall rate in the model is too small by up to 6 mm/day. The general march of the Indian monsoon is well represented by the model, but the northward advance of the major rainband begins nearly 1 month too late.
The simulated differences in precipitation between mid-Holocene and present-day (cf. Sect. 3.1) are smaller than the differences between the simulated present-day climate and the observations. Since the systematic model biases (e.g. biases related to parameterisations) are probably similar in the mid-Holocene and present-day simulations, this fact does not affect the general credibility of the inferred results as long as only differences between the two time-slices are analysed.